风控模型开发验证报告

实验 ID: DEV-mw05p6yt-b571 · 生成时间: 2026-11-03 11:57 (GMT+8) · 报告结构:标准14节模型开发验证模板

00 模型概要信息

项目内容
模型名称exp_DEV-mw05p6yt-b571
开发完成日期2026-11-03
模型开发人员(人工填写)
模型开发人员邮件地址(人工填写)
模型验证人员(人工填写)
模型验证人员邮件地址(人工填写)
模型应用环节collection_scorecard
模型应用频率(人工填写)
模型开发背景及业务用途描述Develop run with tuning
模型适用产品/客户描述(人工填写)
模型重开发原因(如有)—
模型拟监控频率(人工填写)
模型路线LightGBM
目标变量target
实验 IDDEV-mw05p6yt-b571
待补充:模型开发/验证人员、应用频率、产品描述等字段需人工补填。

01 模型设计

建模数据样本窗口

项目内容
切分方式外部 data_flag 划分(TRAIN 内 6/4 随机开发/测试,VALI 跨时间,PSI 时点)
开发样本(Train)窗口—
预留验证样本窗口—
跨时间验证样本窗口—
PSI时点样本窗口—

模型好坏定义

项目内容
目标变量target
坏样本(BAD=1)占比89.72%
好样本(BAD=0)占比—
不确定样本本流程要求目标变量已预计算为 0/1 二值,不确定样本请在上游合并口径

样本排除规则

序号排除规则说明
1id_no主键/时间/口径字段,不入模
2data_mon主键/时间/口径字段,不入模
3data_flag主键/时间/口径字段,不入模
4target 为空排除 0 行

滚动率分析

待补充:滚动率分析(M0→M6+ 状态迁移矩阵)需要逾期状态口径数据,当前流程目标变量为上游已计算的二值标签,暂未生成。
待补充:表现窗口及坏账率分析、模型细分定义需结合业务口径人工补充。

02 数据流

注:中文字典已加载(dictionary.csv,匹配 1 个字段);以下字段未在字典中、报告显示英文原名,请补充字典:id_no、loan_no、data_mon、target、last_con_amt、sum_apply_amt、max_apply_amt、min_apply_amt、avg_apply_amt、valid_sum_apply_amt、last_apply_amt、f_apply_amt、apply_num、max_apprv_amt、min_apprv_amt、avg_apprv_amt、last_apprv_amt、f_apprv_amt、apprv_num、sum_orig_prcp、max_orig_prcp、min_orig_prcp、avg_orig_prcp、last_orig_prcp、f_orig_prcp、loan_num、loan_os_prcp、pp_num、apply_amt_day、apprv_amt_day 等共 584 个。

开发时点 — 样本排除

类别样本数
全部样本13187
排除:目标变量为空0
排除后样本13187

注:按照01模型设计的排除规则统计;当前流程的排除项为目标变量缺失。

总体样本概况

数据集样本量好样本坏样本坏样本率占比
全部样本1318713551183289.72%100.00%
开发样本6858710614889.65%52.01%
测试样本4570472409889.67%34.66%
训练+测试样本1142811821024689.66%86.66%
跨时间验证样本106514092586.85%8.08%
PSI时点样本69433661—5.26%

开发/预留验证样本交叉统计

类别开发+预留样本数开发样本数开发占比预留验证样本数预留验证占比
好118271060.07%47239.93%
坏10246614860.00%409840.00%
合计11428685860.01%457039.99%

细致变量分析(预分箱口径,开发样本)

变量IV缺失率取值集中度最大相关系数
odgt0_m12_num0.34260.00%43.53%0.9372
odgt0_m6_num0.33710.00%45.16%0.9330
odgt0_conti_max_mons_m120.33300.00%50.82%0.9372
ovd_days_d3_month_m60.31920.00%60.22%0.9566
ovd_days_d3_month_m120.31920.00%60.22%0.9566
odgt0_m3_num0.30630.00%48.47%0.9553
odgt0_m6_mob_pc0.30000.00%31.76%0.9052
odgt0_m3_mob_pc0.29050.00%44.91%0.9553
ovd_days_d3_month_m30.28180.00%60.22%0.9102
odgt0_m12_mob_pc0.24340.00%12.72%0.9052
m_c_over_due_cnt0.24270.00%66.64%0.7705
acc_ps_perd_no0.18660.00%25.28%0.8474
ovd_days_d3_count_m120.17490.00%36.56%0.9967
last_od_num_m120.17210.00%37.27%0.9967
ovd_days_d3_count_m60.16890.00%36.66%0.9962
last_od_num_m60.16610.00%37.36%0.9962
ovd_days_d3_count_m30.13090.00%37.49%0.9951
last_od_num_m30.12780.00%38.20%0.9951
all_repay_amount_m3_shd_amt_pc0.12340.00%31.12%0.8383
od_amt_m3_loan_os_prcp_m3_pc0.11570.00%2.04%0.8345
loan_os_prcp0.11460.00%0.38%1.0000
loan_os_prcp_claim0.11460.00%0.38%1.0000
max_loan_os_prcp_m10.11340.00%0.38%0.9998
setl_prcp_ratio0.11200.00%5.29%0.9318
loan_os_prcp_orig_prcp_pc0.10860.00%4.20%0.9318
orig_prcp_avg_m120.10760.00%7.55%0.9497
repay_amount_last_month_loan_os_prcp_pc0.10270.00%65.62%0.5754
balance_avg_m30.10070.00%0.15%1.0000
balance_avg_m3_claim0.10070.00%0.15%1.0000
od_amt_m6_loan_os_prcp_m6_pc0.09960.00%1.95%0.8345
repay_amount_m3_loan_os_prcp_pc0.09690.00%15.16%0.9214
last_loan_setl_tnr0.08930.00%10.73%0.8697
od_amt_loan_os_prcp_pc0.08690.00%8.71%0.9881
ovd_days_max_m120.08650.00%4.78%1.0000
max_od_days_m120.08650.00%4.80%1.0000
max_loan_os_prcp_m30.08640.00%0.41%0.9942
max_od_days0.08490.00%4.77%0.8966
ovd_days_max_m60.08470.00%4.78%1.0000
max_od_days_m60.08470.00%4.80%1.0000
repay_amount_all_shd_amt_pc0.08330.00%63.34%0.7687
ps_od_norm_int_amt0.08220.00%0.09%0.8449
balance_avg_m60.07960.00%0.13%1.0000
balance_avg_m6_claim0.07960.00%0.13%1.0000
avg_od_amt_m3_orig_prcp_pc0.07570.00%26.82%0.9596
ovd_days_max_m30.07510.00%4.84%1.0000
max_od_days_m30.07510.00%4.86%1.0000
max_od_due_m120.07380.00%90.93%0.9982
max_od_due_m60.07200.00%90.94%0.9982
max_od_due0.07180.00%90.86%0.8966
p_mob3_neverdelq_total0.07005.88%65.41%0.4265
apply_amt_avg_m60.06520.00%68.87%0.9529
max_loan_os_prcp_m60.06510.00%1.06%0.9812
apply_amt_avg_m120.06450.00%24.21%0.8431
max_od_due_m30.05970.00%91.34%0.9892
cur_od_loan_amt0.05810.00%1.63%0.8744
avg_od_amt_m6_orig_prcp_pc0.05730.00%21.30%0.9729
all_repay_amount_m6_shd_amt_pc0.05710.00%37.15%0.9488
orig_prcp_avg_m60.05570.00%65.70%0.9529
last_con_amt0.05560.00%6.84%0.9966
max_setl_rem_prcp_paym_m3_last_month_loan_os_prcp_pc0.05500.00%69.55%0.9318
ovd_days_d30_num_m60.05500.00%76.13%0.7081
last_orig_prcp0.05470.00%6.85%0.9966
max_orig_prcp0.05190.00%8.65%0.9185
avg_orig_prcp0.05140.00%2.73%0.9624
last_ovd_to_last_setl_dt_freq0.05111.14%2.35%1.0000
sum_month3_new_loan0.05095.88%21.94%0.9893
all_repay_amount_shd_amt_pc0.05080.00%60.59%0.7687
useratemore90_m6_num0.05020.00%64.25%0.9405
repay_amount_m6_loan_os_prcp_pc0.04950.00%21.77%0.9845
sum_loan_mob24_neverdur0.04935.88%3.86%0.9333
loan_os_prcp_credit_amount_m120.04900.00%1.91%0.3077
loan_os_prcp_credit_amount0.04900.00%15.94%0.9835
od_amt_orig_prcp_pc0.04860.00%31.60%0.8810
loan_os_prcp_credit_amount_m30.04830.00%15.54%0.9835
useratemore50_m3_num_pc0.04790.00%53.31%0.9967
od_amt_m12_loan_os_prcp_m12_pc0.04780.00%2.32%0.8597
loannum0.04785.88%2.22%0.9333
useratemore50monthsubcurmonth0.04760.00%38.54%0.7885
useratemore50_m6_num_pc0.04720.00%38.10%0.9576
useratemore70_m6_num_pc0.04680.00%50.58%0.9529
loan_os_prcp_credit_amount_m60.04680.00%14.63%0.9632
all_repay_amount_m12_shd_amt_pc0.04580.00%44.12%0.9488
balance_avg_m120.04520.00%0.12%1.0000
balance_avg_m12_claim0.04520.00%0.12%1.0000
ps_int_amt_avg0.04480.00%0.04%0.8468
end_loan_to_cur_freq0.04390.00%32.46%0.8135
total_997_cnt_3m0.04374.64%20.67%0.9893
nosvaing_bank_cnt_1m0.04364.64%65.57%0.9962
loanosprcpgt0m12maxcontimons0.04190.00%32.97%0.9924
account_age_loan_day0.04150.00%7.60%0.6156
grxfloanclearednum0.04106.11%4.05%0.8768
max_loan_os_prcp_m120.04000.00%2.61%0.9635
nosvaing_bank_org_1m0.03994.64%65.57%0.9962
nosvaing_bank_rate_1_30.03984.64%34.22%1.0000
useratemore90monthsubcurmonth0.03940.00%40.54%0.7689
account_age0.03840.00%8.15%0.6156
sum_open_loan_mob24_neverdur0.03795.86%17.29%0.6557
useratemore70monthsubcurmonth0.03770.00%28.73%0.7885
t_ps_od_days_max0.03700.00%73.43%0.7435
useratemore50_m12_num_pc0.03640.00%23.78%0.9127

注:共 581 个候选变量,按 IV 降序展示前 100 个;集中度 = 最大频次取值样本数 ÷ 非空样本数;相关性为皮尔逊(仅数值变量),完整数据见 variable_detail.json。

03 单变量分析(细分箱)

odgt0_m12_num — IV 0.3489,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]3004169283594.37%-0.6613
2(1.5, 2.5]1471126134591.43%-0.2092
3(2.5, 3.5]81611070686.52%0.2995
4(3.5, 6.5]114719994882.65%0.5976
5(6.5, inf]42010631474.76%1.0727
odgt0_m6_num — IV 0.3408,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]3117178293994.29%-0.6454
2(1.5, 2.5]1515133138291.22%-0.1823
3(2.5, 3.5]84212272085.51%0.3834
4(3.5, 4.5]56210046282.21%0.6282
5(4.5, inf]82217764578.47%0.8655
odgt0_conti_max_mons_m12 — IV 0.3367,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]3504207329794.09%-0.6094
2(1.5, 2.5]1479150132989.86%-0.0229
3(2.5, 3.5]72412360183.01%0.5722
4(3.5, 4.5]4517237984.04%0.4977
5(4.5, inf]70015854277.43%0.9259
ovd_days_d3_month_m6 — IV 0.3198,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]4427292413593.40%-0.4919
2(1.5, 2.5]92811980987.18%0.2419
3(2.5, 3.5]4998141883.77%0.5176
4(3.5, 5.5]55511344279.64%0.7947
5(5.5, inf]44910534476.61%0.9719
ovd_days_d3_month_m12 — IV 0.3198,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]4427292413593.40%-0.4919
2(1.5, 2.5]92811980987.18%0.2419
3(2.5, 3.5]4998141883.77%0.5176
4(3.5, 5.5]55511344279.64%0.7947
5(5.5, inf]44910534476.61%0.9719
odgt0_m6_mob_pc — IV 0.3159,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 29.17]2641148249394.40%-0.6654
2(29.17, 36.67]1449109134092.48%-0.3505
3(36.67, 63.33]114715299586.75%0.2798
4(63.33, 77.5]62010052083.87%0.5100
5(77.5, inf]100120180079.92%0.7773
ovd_days_d3_month_m3 — IV 0.3097,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]4427292413593.40%-0.4919
2(1.5, 2.5]92811980987.18%0.2419
3(2.5, inf]1503299120480.11%0.7657
odgt0_m3_num — IV 0.3063,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]3344194315094.20%-0.6287
2(1.5, 2.5]1749188156189.25%0.0420
3(2.5, inf]1765328143781.42%0.6813
odgt0_m3_mob_pc — IV 0.2926,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 58.33]3252187306594.25%-0.6381
2(58.33, 83.33]1724186153889.21%0.0461
3(83.33, inf]1882337154582.09%0.6359
odgt0_m12_mob_pc — IV 0.2573,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 23.61]3102188291493.94%-0.5822
2(23.61, 39.75]1652165148790.01%-0.0400
3(39.75, 58.79]119520499182.93%0.5780
4(58.79, 73.87]3945134387.06%0.2527
5(73.87, inf]51510241380.19%0.7601
m_c_over_due_cnt — IV 0.2452,4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]4729343438692.75%-0.3898
2(1.5, 2.5]1179172100785.41%0.3914
3(2.5, 3.5]4518436781.37%0.6841
4(3.5, inf]49911138877.76%0.9071
acc_ps_perd_no — IV 0.2055,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 2.5]3044205283993.27%-0.4696
2(2.5, 5.5]1812197161589.13%0.0547
3(5.5, 6.5]3512332893.45%-0.4989
4(6.5, 16.5]115518397284.16%0.4887
5(16.5, inf]49610239479.44%0.8072
tothismonthodcontiaddnumm12 — IV 0.1837,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]2992445254785.13%0.4140
2(0.5, inf]3866265360193.15%-0.4506
ovd_days_d3_count_m12 — IV 0.1807,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]2805180262593.58%-0.5213
2(1.5, 2.5]1408145126389.70%-0.0059
3(2.5, 3.5]6958261388.20%0.1470
4(3.5, 7.5]1186159102786.59%0.2931
5(7.5, inf]76414462081.15%0.6987
last_od_num_m12 — IV 0.1785,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]2715174254193.59%-0.5226
2(1.5, 2.5]1430144128689.93%-0.0309
3(2.5, 4.5]1236147108988.11%0.1560
4(4.5, 6.5]5537647786.26%0.3218
5(6.5, inf]92416975581.71%0.6618
ovd_days_d3_count_m6 — IV 0.1752,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]2814181263393.57%-0.5188
2(1.5, 2.5]1418149126989.49%0.0166
3(2.5, 5.5]1507202130586.60%0.2929
4(5.5, 7.5]3784133789.15%0.0521
5(7.5, inf]74113760481.51%0.6750
last_od_num_m6 — IV 0.1688,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]2723175254893.57%-0.5197
2(1.5, 2.5]1441147129489.80%-0.0164
3(2.5, 7.5]1934250168487.07%0.2512
4(7.5, 11.5]3667129580.60%0.7343
5(11.5, inf]3946732782.99%0.5734
loan_os_prcp — IV 0.1534,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 396.11]5249542981.87%0.6510
2(396.11, 2274.05]1844217162788.23%0.1440
3(2274.05, 2988.27]4327335983.10%0.5658
4(2988.27, 18390.08]2943269267490.86%-0.1380
5(18390.08, inf]111556105994.98%-0.7811
loan_os_prcp_claim — IV 0.1534,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 396.11]5249542981.87%0.6510
2(396.11, 2274.05]1844217162788.23%0.1440
3(2274.05, 2988.27]4327335983.10%0.5658
4(2988.27, 18390.08]2943269267490.86%-0.1380
5(18390.08, inf]111556105994.98%-0.7811
max_loan_os_prcp_m1 — IV 0.1465,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 396.11]5249542981.87%0.6510
2(396.11, 3956.77]2736338239887.65%0.1993
3(3956.77, 11249.56]1728142158691.78%-0.2545
4(11249.56, 18390.08]7557967689.54%0.0119
5(18390.08, inf]111556105994.98%-0.7811
loan_os_prcp_orig_prcp_pc — IV 0.1437,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 13.04]55011543579.09%0.8282
2(13.04, 53.4]3341372296988.87%0.0815
3(53.4, 67.97]10609796390.85%-0.1367
4(67.97, 76.46]5592853194.99%-0.7839
5(76.46, inf]134898125092.73%-0.3873
ovd_days_d3_count_m3 — IV 0.1350,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]2875191268493.36%-0.4842
2(1.5, 2.5]1510176133488.34%0.1332
3(2.5, 5.5]1507195131287.06%0.2523
4(5.5, 8.5]4847141385.33%0.3978
5(8.5, inf]4827740584.02%0.4985
all_repay_amount_m3_shd_amt_pc — IV 0.1342,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 66.67]3136229290792.70%-0.3825
2(66.67, 66.73]3773833989.92%-0.0298
3(66.73, 69.51]6178453386.39%0.3109
4(69.51, 99.99]69013056081.16%0.6982
5(99.99, inf]2038229180988.76%0.0918
last_od_num_m3 — IV 0.1321,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]2783184259993.39%-0.4893
2(1.5, 2.5]1528175135388.55%0.1133
3(2.5, 4.5]1301172112986.78%0.2770
4(4.5, 6.5]4986243687.55%0.2081
5(6.5, inf]74811763184.36%0.4735
ovd_days_d30_num_m6 — IV 0.1315,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]5221437478491.63%-0.2345
2(0.5, 1.5]112817695284.40%0.4705
3(1.5, inf]5099741280.94%0.7123
od_amt_m3_loan_os_prcp_m3_pc — IV 0.1301,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 10.09]4903415448891.54%-0.2223
2(10.09, 14.35]7168063688.83%0.0854
3(14.35, 24.15]5439345082.87%0.5820
4(24.15, 25.53]3544530987.29%0.2319
5(25.53, inf]3427726577.49%0.9227
repay_amount_last_month_loan_os_prcp_pc — IV 0.1273,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.79]4558382417691.62%-0.2331
2(0.79, 8.66]3445029485.47%0.3871
3(8.66, 20.06]8198073990.23%-0.0647
4(20.06, 49.49]6028851485.38%0.3937
5(49.49, inf]53511042579.44%0.8070
setl_prcp_ratio — IV 0.1266,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.32]1812119169393.43%-0.4965
2(0.32, 0.42]9338285191.21%-0.1811
3(0.42, 0.87]3256354290289.13%0.0548
4(0.87, 1]85715570281.91%0.6481
5(1, inf]0000.00%2.1586
orig_prcp_avg_m12 — IV 0.1218,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 732.05]77914463581.51%0.6748
2(732.05, 1484.13]5987552387.46%0.2165
3(1484.13, 11908.33]3690367332390.05%-0.0446
4(11908.33, 24925]9117383891.99%-0.2819
5(24925, inf]8805182994.20%-0.6298
max_od_days_m12 — IV 0.1213,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4.5]4542842693.83%-0.5636
2(4.5, 7.5]5356646987.66%0.1977
3(7.5, 9.5]4082238694.61%-0.7062
4(9.5, 29.5]4579433414690.54%-0.1005
5(29.5, inf]88216172181.75%0.6594
ovd_days_max_m12 — IV 0.1204,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4.5]4532842593.82%-0.5613
2(4.5, 7.5]5366647087.69%0.1955
3(7.5, 9.5]4062238494.58%-0.7010
4(9.5, 29.5]4580433414790.55%-0.1008
5(29.5, inf]88316172281.77%0.6580
balance_avg_m3 — IV 0.1204,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 380.38]3525929383.24%0.5560
2(380.38, 2752.2]2049244180588.09%0.1575
3(2752.2, 4498.71]87712375485.97%0.3454
4(4498.71, 20051.68]2482228225490.81%-0.1325
5(20051.68, inf]109856104294.90%-0.7649
balance_avg_m3_claim — IV 0.1204,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 380.38]3525929383.24%0.5560
2(380.38, 2752.2]2049244180588.09%0.1575
3(2752.2, 4498.71]87712375485.97%0.3454
4(4498.71, 20051.68]2482228225490.81%-0.1325
5(20051.68, inf]109856104294.90%-0.7649
max_od_days_m6 — IV 0.1199,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4.5]4562842893.86%-0.5683
2(4.5, 7.5]5346646887.64%0.1998
3(7.5, 9.5]4072238594.59%-0.7036
4(9.5, 29.5]4581434414790.53%-0.0985
5(29.5, inf]88016072081.82%0.6545
ovd_days_max_m6 — IV 0.1190,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4.5]4552842793.85%-0.5660
2(4.5, 7.5]5356646987.66%0.1977
3(7.5, 9.5]4052238394.57%-0.6984
4(9.5, 29.5]4582434414890.53%-0.0987
5(29.5, inf]88116072181.84%0.6532
max_od_days — IV 0.1186,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4.5]4512842393.79%-0.5566
2(4.5, 7.5]5376647187.71%0.1934
3(7.5, 9.5]4062238494.58%-0.7010
4(9.5, 29.5]4576433414390.54%-0.0998
5(29.5, inf]88816172781.87%0.6511
ps_od_norm_int_amt — IV 0.1186,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 10.25]4818939281.50%0.6760
2(10.25, 44.34]1489162132789.12%0.0555
3(44.34, 107.69]1553195135887.44%0.2178
4(107.69, 757.16]2829243258691.41%-0.2062
5(757.16, inf]5062148595.85%-0.9810
max_loan_os_prcp_m3 — IV 0.1181,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 533.83]4066733983.50%0.5373
2(533.83, 8748.09]3764442332288.26%0.1416
3(8748.09, 24410.56]1674149152591.10%-0.1672
4(24410.56, 49979.18]6174057793.52%-0.5103
5(49979.18, inf]3971238596.98%-1.3097
od_amt_loan_os_prcp_pc — IV 0.1174,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 8.27]8378275590.20%-0.0614
2(8.27, 12.49]120780112793.37%-0.4867
3(12.49, 47.78]3558362319689.83%-0.0194
4(47.78, 50.31]4002837293.00%-0.4281
5(50.31, inf]85615869881.54%0.6730
od_amt_m6_loan_os_prcp_m6_pc — IV 0.1159,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.36]7667569190.21%-0.0620
2(1.36, 3.57]3875318355791.79%-0.2560
3(3.57, 4.62]7558267389.14%0.0536
4(4.62, 9.86]96313283186.29%0.3188
5(9.86, inf]49910339679.36%0.8119
balance_avg_m6 — IV 0.1135,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 2044.23]1439204123585.82%0.3579
2(2044.23, 9679.95]2833313252088.95%0.0728
3(9679.95, 17757.73]10838599892.15%-0.3045
4(17757.73, 23714.46]4385338587.90%0.1757
5(23714.46, inf]106555101094.84%-0.7518
balance_avg_m6_claim — IV 0.1135,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 2044.23]1439204123585.82%0.3579
2(2044.23, 9679.95]2833313252088.95%0.0728
3(9679.95, 17757.73]10838599892.15%-0.3045
4(17757.73, 23714.46]4385338587.90%0.1757
5(23714.46, inf]106555101094.84%-0.7518
repay_amount_m3_loan_os_prcp_pc — IV 0.1131,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 22.59]1883127175693.26%-0.4680
2(22.59, 51.76]3007296271190.16%-0.0561
3(51.76, 66.23]4406737384.77%0.4417
4(66.23, 99.16]4878839981.93%0.6470
5(99.16, inf]104113290987.32%0.2291
repay_amount_all_shd_amt_pc — IV 0.1084,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 33.16]4535379415691.64%-0.2362
2(33.16, 84.39]3665531184.97%0.4262
3(84.39, 90.67]7709867287.27%0.2333
4(90.67, 96.6]6767859888.46%0.1217
5(96.6, inf]51110041180.43%0.7452
max_od_days_m3 — IV 0.1068,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4.5]4613043193.49%-0.5063
2(4.5, 7.5]5356447188.04%0.1626
3(7.5, 9.5]4082238694.61%-0.7062
4(9.5, 29.5]4603442416190.40%-0.0836
5(29.5, inf]85115269982.14%0.6328
ovd_days_max_m3 — IV 0.1060,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4.5]4603043093.48%-0.5040
2(4.5, 7.5]5366447288.06%0.1605
3(7.5, 9.5]4062238494.58%-0.7010
4(9.5, 29.5]4604442416290.40%-0.0838
5(29.5, inf]85215270082.16%0.6314
end_loan_to_cur_freq — IV 0.1019,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -362.5]4208433680.00%0.7723
2(-362.5, -248.5]1766199156788.73%0.0950
3(-248.5, -109.5]1877159171891.53%-0.2214
4(-109.5, -26.5]5622853495.02%-0.7896
5(-26.5, inf]2233240199389.25%0.0419
all_repay_amount_shd_amt_pc — IV 0.0980,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 3.05]4206353385391.61%-0.2315
2(3.05, 43.26]4746740785.87%0.3545
3(43.26, 50.34]4554940689.23%0.0441
4(50.34, 99.97]3477427378.67%0.8532
5(99.97, inf]1376167120987.86%0.1791
last_loan_setl_tnr — IV 0.0978,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]6454160493.64%-0.5314
2(1.5, 5.5]2418209220991.36%-0.1993
3(5.5, 7.5]1393158123588.66%0.1024
4(7.5, 10.5]1736177155989.80%-0.0170
5(10.5, inf]66612554181.23%0.6935
p_mob3_neverdelq_total — IV 0.0971,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 89.93]5979550284.09%0.4977
2(89.93, 94.01]3687229680.43%0.7487
3(94.01, 96.6]4485139788.62%0.1103
4(96.6, 97.99]3562333393.54%-0.5102
5(97.99, inf]4686425426190.93%-0.1427
max_loan_os_prcp_m6 — IV 0.0965,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 916.64]3495729283.67%0.5249
2(916.64, 1666.65]4053836790.62%-0.1092
3(1666.65, 6861.94]2100271182987.10%0.2492
4(6861.94, 52296.75]3459321313890.72%-0.1213
5(52296.75, inf]5452352295.78%-0.9636
cur_od_loan_amt — IV 0.0936,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 124.01]5728648684.97%0.4268
2(124.01, 558.05]1958227173188.41%0.1271
3(558.05, 725.59]4456837784.72%0.4459
4(725.59, 3545.66]3012283272990.60%-0.1076
5(3545.66, inf]8714682594.72%-0.7281
odduecontireducenumm12 — IV 0.0908,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]5102440466291.38%-0.2018
2(0.5, inf]1756270148684.62%0.4532
odmaxreduceduem12 — IV 0.0908,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]5102440466291.38%-0.2018
2(0.5, inf]1756270148684.62%0.4532
loan_os_prcp_credit_amount — IV 0.0907,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -4998.92]109312097389.02%0.0657
2(-4998.92, 9.88]77212964383.29%0.5523
3(9.88, 17.78]5526948387.50%0.2127
4(17.78, 52.45]1949208174189.33%0.0339
5(52.45, inf]2492184230892.62%-0.3706
max_setl_rem_prcp_paym_m3_last_month_loan_os_prcp_pc — IV 0.0888,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.88]4795544425188.65%0.1027
2(1.88, 11.68]5442052496.32%-1.1071
3(11.68, 20.03]6655860791.28%-0.1895
4(20.03, 33.46]3582433493.30%-0.4745
5(33.46, inf]4966443287.10%0.2491
avg_od_amt_m3_orig_prcp_pc — IV 0.0878,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 2.22]5507048087.27%0.2333
2(2.22, 3.04]4069337373291.72%-0.2460
3(3.04, 5.76]1271148112388.36%0.1321
4(5.76, 9.35]5599946082.29%0.6225
5(9.35, inf]4095635386.31%0.3175
all_repay_amount_m6_shd_amt_pc — IV 0.0875,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 75.01]8946183393.18%-0.4555
2(75.01, 83.35]2432210222291.37%-0.2004
3(83.35, 83.65]4556539085.71%0.3669
4(83.65, 99.56]5249443082.06%0.6381
5(99.56, inf]2553280227389.03%0.0646
all_repay_amount_m12_shd_amt_pc — IV 0.0849,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 83.35]1564107145793.16%-0.4527
2(83.35, 89.74]1218140107888.51%0.1174
3(89.74, 91.06]5615051191.09%-0.1657
4(91.06, 99.75]4858939681.65%0.6658
5(99.75, inf]3030324270689.31%0.0361
loan_os_prcp_credit_amount_m6 — IV 0.0834,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 2.45]104911593489.04%0.0641
2(2.45, 24.34]106115690585.30%0.4005
3(24.34, 65.7]2221248197388.83%0.0847
4(65.7, 113.44]2140148199293.08%-0.4411
5(113.44, inf]3874334488.89%0.0792
max_orig_prcp — IV 0.0800,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1850]6759757885.63%0.3738
2(1850, 11050]3580406317488.66%0.1022
3(11050, 30650]1619128149192.09%-0.2966
4(30650, 47050]3704732387.30%0.2311
5(47050, inf]6143258294.79%-0.7421
balance_avg_m12 — IV 0.0799,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4286.8]2115275184087.00%0.2579
2(4286.8, 5786.94]6335957490.68%-0.1165
3(5786.94, 6764.49]3685031886.41%0.3086
4(6764.49, 48615.01]3314309300590.68%-0.1161
5(48615.01, inf]4281741196.03%-1.0268
balance_avg_m12_claim — IV 0.0799,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4286.8]2115275184087.00%0.2579
2(4286.8, 5786.94]6335957490.68%-0.1165
3(5786.94, 6764.49]3685031886.41%0.3086
4(6764.49, 48615.01]3314309300590.68%-0.1161
5(48615.01, inf]4281741196.03%-1.0268
apply_amt_avg_m12 — IV 0.0793,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4525]3004381262387.32%0.2293
2(4525, 10810]1167124104389.37%0.0290
3(10810, 32366.67]1481110137192.57%-0.3642
4(32366.67, 48025]5025344989.44%0.0219
5(48025, inf]7044266294.03%-0.5990
ovd_days_d30_count_m12 — IV 0.0772,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]6258591566790.56%-0.1020
2(0.5, inf]60011948180.17%0.7619
orig_prcp_avg_m6 — IV 0.0769,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 636.67]4606542406488.23%0.1440
2(636.67, 2158.33]4704642490.21%-0.0625
3(2158.33, 5016.67]5274748091.08%-0.1650
4(5016.67, 15012.5]6314658592.71%-0.3844
5(15012.5, inf]6242959595.35%-0.8626
ps_int_amt_avg — IV 0.0765,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 12.24]6468656086.69%0.2850
2(12.24, 17.12]4647439084.05%0.4965
3(17.12, 22.93]4573442392.56%-0.3624
4(22.93, 114.22]3253366288788.75%0.0933
5(114.22, inf]2038150188892.64%-0.3740
apply_amt_avg_m6 — IV 0.0759,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 636.67]4787562422588.26%0.1413
2(636.67, 4525]6566259490.55%-0.1011
3(4525, 15100]6654562093.23%-0.4644
4(15100, 32314.29]3581634295.53%-0.9036
5(32314.29, inf]3922536793.62%-0.5279
m1_ovd_days_num — IV 0.0755,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]6253591566290.55%-0.1011
2(0.5, inf]60511948680.33%0.7515
ovd_days_d30_month_m3 — IV 0.0755,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]6253591566290.55%-0.1011
2(0.5, inf]60511948680.33%0.7515
ovd_days_d30_month_m6 — IV 0.0755,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]6253591566290.55%-0.1011
2(0.5, inf]60511948680.33%0.7515
ovd_days_d30_month_m12 — IV 0.0755,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]6253591566290.55%-0.1011
2(0.5, inf]60511948680.33%0.7515
ovd_days_d30_count_m6 — IV 0.0754,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]6260592566890.54%-0.1005
2(0.5, inf]59811848080.27%0.7555
avg_od_amt_m6_orig_prcp_pc — IV 0.0751,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.3]5746650888.50%0.1178
2(1.3, 2.9]4994449454591.01%-0.1561
3(2.9, 4.18]5947951586.70%0.2839
4(4.18, 6.97]3467227479.19%0.8222
5(6.97, inf]3504430687.43%0.2192
odgt1_m12_mob_pc — IV 0.0738,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4.16]6255592566390.54%-0.0996
2(4.16, inf]60311848580.43%0.7452
odgt1_m12_num — IV 0.0738,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]6255592566390.54%-0.0996
2(0.5, inf]60311848580.43%0.7452
odgt1_m12_subcurmonth — IV 0.0738,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 50003]60311848580.43%0.7452
2(50003, inf]6255592566390.54%-0.0996
max_od_due_m12 — IV 0.0738,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]6255592566390.54%-0.0996
2(1.5, inf]60311848580.43%0.7452
odgt1_conti_max_mons_m12 — IV 0.0738,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]6255592566390.54%-0.0996
2(0.5, inf]60311848580.43%0.7452
account_age_loan_day — IV 0.0736,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.52]1645118152792.83%-0.4018
2(0.52, 0.96]2199240195989.09%0.0591
3(0.96, 1.19]64610753983.44%0.5417
4(1.19, 4.62]1453137131690.57%-0.1038
5(4.62, inf]91510880788.20%0.1474
repay_amount_m6_loan_os_prcp_pc — IV 0.0729,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 42.01]2194173202192.11%-0.2994
2(42.01, 70.26]2298228207090.08%-0.0473
3(70.26, 82.59]5236945486.81%0.2746
4(82.59, 99.67]3496828180.52%0.7398
5(99.67, inf]1494172132288.49%0.1192
sum_orig_prcp — IV 0.0720,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 10450]2505314219187.47%0.2159
2(10450, 12554]3552133494.08%-0.6080
3(12554, 19150]8259672988.36%0.1313
4(19150, 110100]2809263254690.64%-0.1115
5(110100, inf]3641634895.60%-0.9210
odgt1_m6_mob_pc — IV 0.0720,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 8.34]6257593566490.52%-0.0981
2(8.34, inf]60111748480.53%0.7387
odgt1_m6_num — IV 0.0720,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]6257593566490.52%-0.0981
2(0.5, inf]60111748480.53%0.7387
max_od_due_m6 — IV 0.0720,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]6257593566490.52%-0.0981
2(1.5, inf]60111748480.53%0.7387
max_od_due — IV 0.0718,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]6249592565790.53%-0.0985
2(1.5, inf]60911849180.62%0.7329
loan_os_prcp_credit_amount_m3 — IV 0.0716,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.22]109811997989.16%0.0512
2(1.22, 8.26]4317535682.60%0.6012
3(8.26, 18.51]6708958186.72%0.2825
4(18.51, 51.44]1860206165488.92%0.0755
5(51.44, inf]2799221257892.10%-0.2980
loannum — IV 0.0712,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 9.5]4357036583.91%0.5110
2(9.5, 27.5]2020242177888.02%0.1681
3(27.5, 39.5]10388795191.62%-0.2292
4(39.5, 61.5]1212137107588.70%0.1023
5(61.5, inf]1750130162092.57%-0.3602
sum_loan_mob24_neverdur — IV 0.0705,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 10.5]2290296199487.07%0.2549
2(10.5, 22.5]1485155133089.56%0.0129
3(22.5, 27.5]3712135094.34%-0.6510
4(27.5, 43.5]8959080589.94%-0.0286
5(43.5, inf]1414104131092.64%-0.3710
last_ovd_to_last_setl_dt_freq — IV 0.0662,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 22.5]4842945594.01%-0.5929
2(22.5, 201.5]2476281219588.65%0.1045
3(201.5, 352.5]2532221231191.27%-0.1872
4(352.5, 390.5]92110481788.71%0.0988
5(390.5, inf]3676630182.02%0.6426
ovd_days_d30_count_m3 — IV 0.0659,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]6289600568990.46%-0.0907
2(0.5, inf]56911045980.67%0.7300
sum_card_mob12_neverdur — IV 0.0657,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]2098252184687.99%0.1711
2(0.5, 1.5]8347775790.77%-0.1231
3(1.5, 6.5]1857214164388.48%0.1241
4(6.5, 8.5]3351332296.12%-1.0472
5(8.5, inf]1331110122191.74%-0.2445
last_con_amt — IV 0.0656,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1150]114915999086.16%0.3298
2(1150, 2950]10809798391.02%-0.1573
3(2950, 7850]1739211152887.87%0.1788
4(7850, 47850]2435218221791.05%-0.1608
5(47850, inf]4552543094.51%-0.6863
od_amt_m12_loan_os_prcp_m12_pc — IV 0.0654,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.91]1241129111289.61%0.0045
2(0.91, 1.09]3855133486.75%0.2793
3(1.09, 2.17]3540299324191.55%-0.2246
4(2.17, 3.17]7828969388.62%0.1062
5(3.17, inf]91014276884.40%0.4707
setl_int_mon — IV 0.0652,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 3.64]5009457455290.88%-0.1400
2(3.64, 17.71]3445928582.85%0.5837
3(17.71, 42.33]3704632487.57%0.2065
4(42.33, 90.6]3776731082.23%0.6267
5(90.6, inf]7588167789.31%0.0354
last_orig_prcp — IV 0.0651,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1150]115015999186.17%0.3288
2(1150, 2950]10779798090.99%-0.1542
3(2950, 7850]1739211152887.87%0.1788
4(7850, 47850]2439218222191.06%-0.1626
5(47850, inf]4532542894.48%-0.6816
repay_amount_m3_all_shd_amt_pc — IV 0.0651,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 82.04]4552942693.63%-0.5285
2(82.04, 86.14]5445349190.26%-0.0675
3(86.14, 87.85]3634831586.78%0.2772
4(87.85, 94.69]3819345347490.97%-0.1509
5(94.69, inf]1677235144285.99%0.3444
setl_rem_prcp_paym_m3_loan_os_prcp_pc — IV 0.0649,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -4999]3472232593.66%-0.5342
2(-4999, 3.21]4506525398188.35%0.1327
3(3.21, 19.05]8575080794.17%-0.6227
4(19.05, 25.49]3654232388.49%0.1186
5(25.49, inf]7837171290.93%-0.1468
avg_orig_prcp — IV 0.0647,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 876.79]3585830083.80%0.5153
2(876.79, 1493.1]5978151686.43%0.3070
3(1493.1, 10007.14]3924419350589.32%0.0345
4(10007.14, 42550]1566131143591.63%-0.2351
5(42550, inf]4132139294.92%-0.7681
max_loan_os_prcp_m12 — IV 0.0642,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1305.9]3435029385.42%0.3905
2(1305.9, 5100.68]1326155117188.31%0.1364
3(5100.68, 7500.04]5558047585.59%0.3773
4(7500.04, 63002.38]4096398369890.28%-0.0705
5(63002.38, inf]5382751194.98%-0.7819

注:统计全部候选变量的自动分箱明细(初分组),按 IV 降序展示前 100 个变量,共 568 个;完整数据见 feature_report.json。

04 单变量分析(粗分箱)

odgt0_m12_num — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]3004169283594.37%-0.6613
2(1.5, 2.5]1471126134591.43%-0.2092
3(2.5, 3.5]81611070686.52%0.2995
4(3.5, 6.5]114719994882.65%0.5976
5(6.5, inf]42010631474.76%1.0727
2026-11-03T11:57:40.388084 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调下降(94.37% → 74.76%),WOE 单调上升(-0.6613 → 1.0727),符合风控分箱预期。IV=0.3489,该变量区分度尚可。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 1 箱((-inf, 1.5])为极高风险区间:坏率高达 94.37%,策略上应重点管控

odgt0_m12_mob_pc — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 23.61]3102188291493.94%-0.5822
2(23.61, 39.75]1652165148790.01%-0.0400
3(39.75, 58.79]119520499182.93%0.5780
4(58.79, 73.87]3945134387.06%0.2527
5(73.87, inf]51510241380.19%0.7601
2026-11-03T11:57:40.522768 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(93.94% → 93.94% → 80.19%),如需严格单调可考虑调整切点或合并分箱。IV=0.2573,该变量区分度尚可。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 82.93% 回升至 87.06%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 23.61])为极高风险区间:坏率高达 93.94%,策略上应重点管控

m_c_over_due_cnt — 4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]4729343438692.75%-0.3898
2(1.5, 2.5]1179172100785.41%0.3914
3(2.5, 3.5]4518436781.37%0.6841
4(3.5, inf]49911138877.76%0.9071
2026-11-03T11:57:40.691208 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调下降(92.75% → 77.76%),WOE 单调上升(-0.3898 → 0.9071),符合风控分箱预期。IV=0.2452,该变量区分度尚可。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 1 箱((-inf, 1.5])为极高风险区间:坏率高达 92.75%,策略上应重点管控

acc_ps_perd_no — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 2.5]3044205283993.27%-0.4696
2(2.5, 5.5]1812197161589.13%0.0547
3(5.5, 6.5]3512332893.45%-0.4989
4(6.5, 16.5]115518397284.16%0.4887
5(16.5, inf]49610239479.44%0.8072
2026-11-03T11:57:40.821409 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(93.27% → 93.45% → 79.44%),如需严格单调可考虑调整切点或合并分箱。IV=0.2055,该变量区分度尚可。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 89.13% 回升至 93.45%,如需严格单调可考虑合并这两箱

第 3 箱((5.5, 6.5])为极高风险区间:坏率高达 93.45%,策略上应重点管控

tothismonthodcontiaddnumm12 — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]2992445254785.13%0.4140
2(0.5, inf]3866265360193.15%-0.4506
2026-11-03T11:57:40.938590 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调上升(85.13% → 93.15%),WOE 单调下降(0.4140 → -0.4506),符合风控分箱预期。IV=0.1837,该变量区分度尚可。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 1 箱 WOE<0(风险高于平均)

第 2 箱((0.5, inf])为极高风险区间:坏率高达 93.15%,策略上应重点管控

loan_os_prcp_orig_prcp_pc — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 13.04]55011543579.09%0.8282
2(13.04, 53.4]3341372296988.87%0.0815
3(53.4, 67.97]10609796390.85%-0.1367
4(67.97, 76.46]5592853194.99%-0.7839
5(76.46, inf]134898125092.73%-0.3873
2026-11-03T11:57:41.062804 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(79.09% → 94.99% → 92.73%),如需严格单调可考虑调整切点或合并分箱。IV=0.1437,该变量区分度尚可。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 4→5 箱有局部反弹:Bad Rate 由 94.99% 回落至 92.73%,如需严格单调可考虑合并这两箱

第 4 箱((67.97, 76.46])为极高风险区间:坏率高达 94.99%,策略上应重点管控

all_repay_amount_m3_shd_amt_pc — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 66.67]3136229290792.70%-0.3825
2(66.67, 66.73]3773833989.92%-0.0298
3(66.73, 69.51]6178453386.39%0.3109
4(69.51, 99.99]69013056081.16%0.6982
5(99.99, inf]2038229180988.76%0.0918
2026-11-03T11:57:41.190276 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(92.70% → 92.70% → 88.76%),如需严格单调可考虑调整切点或合并分箱。IV=0.1342,该变量区分度尚可。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 4→5 箱有局部反弹:Bad Rate 由 81.16% 回升至 88.76%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 66.67])为极高风险区间:坏率高达 92.70%,策略上应重点管控

ovd_days_d30_num_m6 — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]5221437478491.63%-0.2345
2(0.5, 1.5]112817695284.40%0.4705
3(1.5, inf]5099741280.94%0.7123
2026-11-03T11:57:41.303949 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调下降(91.63% → 80.94%),WOE 单调上升(-0.2345 → 0.7123),符合风控分箱预期。IV=0.1315,该变量区分度尚可。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 1 箱((-inf, 0.5])为极高风险区间:坏率高达 91.63%,策略上应重点管控

od_amt_m3_loan_os_prcp_m3_pc — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 10.09]4903415448891.54%-0.2223
2(10.09, 14.35]7168063688.83%0.0854
3(14.35, 24.15]5439345082.87%0.5820
4(24.15, 25.53]3544530987.29%0.2319
5(25.53, inf]3427726577.49%0.9227
2026-11-03T11:57:41.422172 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(91.54% → 91.54% → 77.49%),如需严格单调可考虑调整切点或合并分箱。IV=0.1301,该变量区分度尚可。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 82.87% 回升至 87.29%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 10.09])为极高风险区间:坏率高达 91.54%,策略上应重点管控

repay_amount_last_month_loan_os_prcp_pc — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.79]4558382417691.62%-0.2331
2(0.79, 8.66]3445029485.47%0.3871
3(8.66, 20.06]8198073990.23%-0.0647
4(20.06, 49.49]6028851485.38%0.3937
5(49.49, inf]53511042579.44%0.8070
2026-11-03T11:57:41.548689 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(91.62% → 91.62% → 79.44%),如需严格单调可考虑调整切点或合并分箱。IV=0.1273,该变量区分度尚可。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 85.47% 回升至 90.23%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 0.79])为极高风险区间:坏率高达 91.62%,策略上应重点管控

max_od_days_m12 — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4.5]4542842693.83%-0.5636
2(4.5, 7.5]5356646987.66%0.1977
3(7.5, 9.5]4082238694.61%-0.7062
4(9.5, 29.5]4579433414690.54%-0.1005
5(29.5, inf]88216172181.75%0.6594
2026-11-03T11:57:41.675589 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(93.83% → 94.61% → 81.75%),如需严格单调可考虑调整切点或合并分箱。IV=0.1213,该变量区分度尚可。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 87.66% 回升至 94.61%,如需严格单调可考虑合并这两箱

第 3 箱((7.5, 9.5])为极高风险区间:坏率高达 94.61%,策略上应重点管控

ps_od_norm_int_amt — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 10.25]4818939281.50%0.6760
2(10.25, 44.34]1489162132789.12%0.0555
3(44.34, 107.69]1553195135887.44%0.2178
4(107.69, 757.16]2829243258691.41%-0.2062
5(757.16, inf]5062148595.85%-0.9810
2026-11-03T11:57:41.805429 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(81.50% → 95.85% → 95.85%),如需严格单调可考虑调整切点或合并分箱。IV=0.1186,该变量区分度尚可。

WOE 在第 3、4 箱之间穿越 0 线:前 3 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 89.12% 回落至 87.44%,如需严格单调可考虑合并这两箱

第 5 箱((757.16, inf])为极高风险区间:坏率高达 95.85%,策略上应重点管控

repay_amount_all_shd_amt_pc — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 33.16]4535379415691.64%-0.2362
2(33.16, 84.39]3665531184.97%0.4262
3(84.39, 90.67]7709867287.27%0.2333
4(90.67, 96.6]6767859888.46%0.1217
5(96.6, inf]51110041180.43%0.7452
2026-11-03T11:57:41.933451 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(91.64% → 91.64% → 80.43%),如需严格单调可考虑调整切点或合并分箱。IV=0.1084,该变量区分度尚可。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 84.97% 回升至 87.27%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 33.16])为极高风险区间:坏率高达 91.64%,策略上应重点管控

end_loan_to_cur_freq — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -362.5]4208433680.00%0.7723
2(-362.5, -248.5]1766199156788.73%0.0950
3(-248.5, -109.5]1877159171891.53%-0.2214
4(-109.5, -26.5]5622853495.02%-0.7896
5(-26.5, inf]2233240199389.25%0.0419
2026-11-03T11:57:42.128043 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(80.00% → 95.02% → 89.25%),如需严格单调可考虑调整切点或合并分箱。IV=0.1019,该变量区分度尚可。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 4→5 箱有局部反弹:Bad Rate 由 95.02% 回落至 89.25%,如需严格单调可考虑合并这两箱

第 4 箱((-109.5, -26.5])为极高风险区间:坏率高达 95.02%,策略上应重点管控

cur_od_loan_amt — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 124.01]5728648684.97%0.4268
2(124.01, 558.05]1958227173188.41%0.1271
3(558.05, 725.59]4456837784.72%0.4459
4(725.59, 3545.66]3012283272990.60%-0.1076
5(3545.66, inf]8714682594.72%-0.7281
2026-11-03T11:57:42.253802 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(84.97% → 94.72% → 94.72%),如需严格单调可考虑调整切点或合并分箱。IV=0.0936,该变量区分度偏弱但可用。

WOE 在第 3、4 箱之间穿越 0 线:前 3 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 88.41% 回落至 84.72%,如需严格单调可考虑合并这两箱

第 5 箱((3545.66, inf])为极高风险区间:坏率高达 94.72%,策略上应重点管控

odduecontireducenumm12 — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]5102440466291.38%-0.2018
2(0.5, inf]1756270148684.62%0.4532
2026-11-03T11:57:42.367015 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调下降(91.38% → 84.62%),WOE 单调上升(-0.2018 → 0.4532),符合风控分箱预期。IV=0.0908,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 1 箱 WOE<0(风险高于平均)

第 1 箱((-inf, 0.5])为极高风险区间:坏率高达 91.38%,策略上应重点管控

max_setl_rem_prcp_paym_m3_last_month_loan_os_prcp_pc — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.88]4795544425188.65%0.1027
2(1.88, 11.68]5442052496.32%-1.1071
3(11.68, 20.03]6655860791.28%-0.1895
4(20.03, 33.46]3582433493.30%-0.4745
5(33.46, inf]4966443287.10%0.2491
2026-11-03T11:57:42.482400 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.65% → 96.32% → 87.10%),如需严格单调可考虑调整切点或合并分箱。IV=0.0888,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 88.65% 回升至 96.32%,如需严格单调可考虑合并这两箱

第 2 箱((1.88, 11.68])为极高风险区间:坏率高达 96.32%,策略上应重点管控

apply_amt_avg_m12 — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4525]3004381262387.32%0.2293
2(4525, 10810]1167124104389.37%0.0290
3(10810, 32366.67]1481110137192.57%-0.3642
4(32366.67, 48025]5025344989.44%0.0219
5(48025, inf]7044266294.03%-0.5990
2026-11-03T11:57:42.607926 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.32% → 94.03% → 94.03%),如需严格单调可考虑调整切点或合并分箱。IV=0.0793,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 92.57% 回落至 89.44%,如需严格单调可考虑合并这两箱

第 5 箱((48025, inf])为极高风险区间:坏率高达 94.03%,策略上应重点管控

orig_prcp_avg_m6 — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 636.67]4606542406488.23%0.1440
2(636.67, 2158.33]4704642490.21%-0.0625
3(2158.33, 5016.67]5274748091.08%-0.1650
4(5016.67, 15012.5]6314658592.71%-0.3844
5(15012.5, inf]6242959595.35%-0.8626
2026-11-03T11:57:42.732521 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调上升(88.23% → 95.35%),WOE 单调下降(0.1440 → -0.8626),符合风控分箱预期。IV=0.0769,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 5 箱((15012.5, inf])为极高风险区间:坏率高达 95.35%,策略上应重点管控

account_age_loan_day — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.52]1645118152792.83%-0.4018
2(0.52, 0.96]2199240195989.09%0.0591
3(0.96, 1.19]64610753983.44%0.5417
4(1.19, 4.62]1453137131690.57%-0.1038
5(4.62, inf]91510880788.20%0.1474
2026-11-03T11:57:42.852534 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(92.83% → 92.83% → 88.20%),如需严格单调可考虑调整切点或合并分箱。IV=0.0736,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 83.44% 回升至 90.57%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 0.52])为极高风险区间:坏率高达 92.83%,策略上应重点管控

repay_amount_m6_loan_os_prcp_pc — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 42.01]2194173202192.11%-0.2994
2(42.01, 70.26]2298228207090.08%-0.0473
3(70.26, 82.59]5236945486.81%0.2746
4(82.59, 99.67]3496828180.52%0.7398
5(99.67, inf]1494172132288.49%0.1192
2026-11-03T11:57:42.971692 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(92.11% → 92.11% → 88.49%),如需严格单调可考虑调整切点或合并分箱。IV=0.0729,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 4→5 箱有局部反弹:Bad Rate 由 80.52% 回升至 88.49%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 42.01])为极高风险区间:坏率高达 92.11%,策略上应重点管控

sum_orig_prcp — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 10450]2505314219187.47%0.2159
2(10450, 12554]3552133494.08%-0.6080
3(12554, 19150]8259672988.36%0.1313
4(19150, 110100]2809263254690.64%-0.1115
5(110100, inf]3641634895.60%-0.9210
2026-11-03T11:57:43.097861 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.47% → 95.60% → 95.60%),如需严格单调可考虑调整切点或合并分箱。IV=0.0720,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 94.08% 回落至 88.36%,如需严格单调可考虑合并这两箱

第 5 箱((110100, inf])为极高风险区间:坏率高达 95.60%,策略上应重点管控

loannum — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 9.5]4357036583.91%0.5110
2(9.5, 27.5]2020242177888.02%0.1681
3(27.5, 39.5]10388795191.62%-0.2292
4(39.5, 61.5]1212137107588.70%0.1023
5(61.5, inf]1750130162092.57%-0.3602
2026-11-03T11:57:43.219050 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(83.91% → 92.57% → 92.57%),如需严格单调可考虑调整切点或合并分箱。IV=0.0712,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 91.62% 回落至 88.70%,如需严格单调可考虑合并这两箱

第 5 箱((61.5, inf])为极高风险区间:坏率高达 92.57%,策略上应重点管控

sum_card_mob12_neverdur — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]2098252184687.99%0.1711
2(0.5, 1.5]8347775790.77%-0.1231
3(1.5, 6.5]1857214164388.48%0.1241
4(6.5, 8.5]3351332296.12%-1.0472
5(8.5, inf]1331110122191.74%-0.2445
2026-11-03T11:57:43.407269 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.99% → 96.12% → 91.74%),如需严格单调可考虑调整切点或合并分箱。IV=0.0657,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 90.77% 回落至 88.48%,如需严格单调可考虑合并这两箱

第 4 箱((6.5, 8.5])为极高风险区间:坏率高达 96.12%,策略上应重点管控

od_amt_m12_loan_os_prcp_m12_pc — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.91]1241129111289.61%0.0045
2(0.91, 1.09]3855133486.75%0.2793
3(1.09, 2.17]3540299324191.55%-0.2246
4(2.17, 3.17]7828969388.62%0.1062
5(3.17, inf]91014276884.40%0.4707
2026-11-03T11:57:43.538188 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(89.61% → 91.55% → 84.40%),如需严格单调可考虑调整切点或合并分箱。IV=0.0654,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 86.75% 回升至 91.55%,如需严格单调可考虑合并这两箱

第 3 箱((1.09, 2.17])为极高风险区间:坏率高达 91.55%,策略上应重点管控

setl_int_mon — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 3.64]5009457455290.88%-0.1400
2(3.64, 17.71]3445928582.85%0.5837
3(17.71, 42.33]3704632487.57%0.2065
4(42.33, 90.6]3776731082.23%0.6267
5(90.6, inf]7588167789.31%0.0354
2026-11-03T11:57:43.664431 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(90.88% → 90.88% → 89.31%),如需严格单调可考虑调整切点或合并分箱。IV=0.0652,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 82.85% 回升至 87.57%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 3.64])为极高风险区间:坏率高达 90.88%,策略上应重点管控

setl_rem_prcp_paym_m3_loan_os_prcp_pc — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -4999]3472232593.66%-0.5342
2(-4999, 3.21]4506525398188.35%0.1327
3(3.21, 19.05]8575080794.17%-0.6227
4(19.05, 25.49]3654232388.49%0.1186
5(25.49, inf]7837171290.93%-0.1468
2026-11-03T11:57:43.789425 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(93.66% → 94.17% → 90.93%),如需严格单调可考虑调整切点或合并分箱。IV=0.0649,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 88.35% 回升至 94.17%,如需严格单调可考虑合并这两箱

第 3 箱((3.21, 19.05])为极高风险区间:坏率高达 94.17%,策略上应重点管控

sum_month3_new_loan — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]1416182123487.15%0.2484
2(0.5, 1.5]1328142118689.31%0.0399
3(1.5, 2.5]108813695287.50%0.2165
4(2.5, 3.5]7565270493.12%-0.4431
5(3.5, inf]1867154171391.75%-0.2466
2026-11-03T11:57:43.916094 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.15% → 93.12% → 91.75%),如需严格单调可考虑调整切点或合并分箱。IV=0.0591,该变量区分度偏弱但可用。

WOE 在第 3、4 箱之间穿越 0 线:前 3 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 89.31% 回落至 87.50%,如需严格单调可考虑合并这两箱

第 4 箱((2.5, 3.5])为极高风险区间:坏率高达 93.12%,策略上应重点管控

sum_apply_amt — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 2050]5137543885.38%0.3939
2(2050, 9625]1776187158989.47%0.0189
3(9625, 11585]3906332783.85%0.5118
4(11585, 130400]3817366345190.41%-0.0852
5(130400, inf]3621934394.75%-0.7347
2026-11-03T11:57:44.038531 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(85.38% → 94.75% → 94.75%),如需严格单调可考虑调整切点或合并分箱。IV=0.0571,该变量区分度偏弱但可用。

WOE 在第 3、4 箱之间穿越 0 线:前 3 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 89.47% 回落至 83.85%,如需严格单调可考虑合并这两箱

第 5 箱((130400, inf])为极高风险区间:坏率高达 94.75%,策略上应重点管控

loan_reject_rate_3m — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -0.15]8256576092.12%-0.2966
2(-0.15, 0.02]1520172134888.68%0.1034
3(0.02, 0.28]4632843593.95%-0.5808
4(0.28, 0.58]1341115122691.42%-0.2043
5(0.58, inf]2286284200287.58%0.2094
2026-11-03T11:57:44.165982 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(92.12% → 93.95% → 87.58%),如需严格单调可考虑调整切点或合并分箱。IV=0.0569,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 88.68% 回升至 93.95%,如需严格单调可考虑合并这两箱

第 3 箱((0.02, 0.28])为极高风险区间:坏率高达 93.95%,策略上应重点管控

setl_prcp_mon_ratio — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0]4990458453290.82%-0.1334
2(0, 0.01]5096944086.44%0.3059
3(0.01, 0.02]3733334091.15%-0.1738
4(0.02, 0.04]4727639683.90%0.5079
5(0.04, inf]5147444085.60%0.3759
2026-11-03T11:57:44.297726 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(90.82% → 91.15% → 85.60%),如需严格单调可考虑调整切点或合并分箱。IV=0.0556,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 86.44% 回升至 91.15%,如需严格单调可考虑合并这两箱

第 3 箱((0.01, 0.02])为极高风险区间:坏率高达 91.15%,策略上应重点管控

useratemore90_m6_num — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]4406515389188.31%0.1364
2(0.5, 1.5]5705351790.70%-0.1191
3(1.5, 2.5]8605380793.84%-0.5644
4(2.5, 3.5]4413141092.97%-0.4236
5(3.5, inf]5815852390.02%-0.0405
2026-11-03T11:57:44.423889 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.31% → 93.84% → 90.02%),如需严格单调可考虑调整切点或合并分箱。IV=0.0556,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 93.84% 回落至 92.97%,如需严格单调可考虑合并这两箱

第 3 箱((1.5, 2.5])为极高风险区间:坏率高达 93.84%,策略上应重点管控

account_age — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 6.5]1897151174692.04%-0.2892
2(6.5, 11.5]2514270224489.26%0.0410
3(11.5, 14.5]71711360484.24%0.4824
4(14.5, 33.5]1350146120489.19%0.0488
5(33.5, inf]3803035092.11%-0.2981
2026-11-03T11:57:44.555190 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(92.04% → 92.11% → 92.11%),如需严格单调可考虑调整切点或合并分箱。IV=0.0555,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 92.04% 回落至 89.26%,如需严格单调可考虑合并这两箱

第 5 箱((33.5, inf])为极高风险区间:坏率高达 92.11%,策略上应重点管控

apply_num_m6 — 4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]4480531394988.15%0.1522
2(0.5, 1.5]1439112132792.22%-0.3136
3(1.5, 2.5]4643742792.03%-0.2872
4(2.5, inf]4753044593.68%-0.5383
2026-11-03T11:57:44.681644 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.15% → 93.68% → 93.68%),如需严格单调可考虑调整切点或合并分箱。IV=0.0555,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 92.22% 回落至 92.03%,如需严格单调可考虑合并这两箱

第 4 箱((2.5, inf])为极高风险区间:坏率高达 93.68%,策略上应重点管控

repay_amount_m12_all_shd_amt_pc — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 85.33]1183111107290.62%-0.1091
2(85.33, 86.76]3742235294.12%-0.6140
3(86.76, 89.01]4798139883.09%0.5666
4(89.01, 99.76]4478471400789.48%0.0177
5(99.76, inf]3442531992.73%-0.3877
2026-11-03T11:57:44.873573 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(90.62% → 94.12% → 92.73%),如需严格单调可考虑调整切点或合并分箱。IV=0.0527,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 94.12% 回落至 83.09%,如需严格单调可考虑合并这两箱

第 2 箱((85.33, 86.76])为极高风险区间:坏率高达 94.12%,策略上应重点管控

t_ps_od_days_max — 4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]5036466457090.75%-0.1245
2(0.5, 3.5]77110966285.86%0.3547
3(3.5, 11.5]4327335983.10%0.5658
4(11.5, inf]6196255789.98%-0.0368
2026-11-03T11:57:45.001873 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(90.75% → 90.75% → 89.98%),如需严格单调可考虑调整切点或合并分箱。IV=0.0523,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 83.10% 回升至 89.98%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 0.5])为极高风险区间:坏率高达 90.75%,策略上应重点管控

nosvaing_bank_cnt_1m — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]4288492379688.53%0.1188
2(0.5, 1.5]1421135128690.50%-0.0920
3(1.5, inf]8314878394.22%-0.6299
2026-11-03T11:57:45.122071 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调上升(88.53% → 94.22%),WOE 单调下降(0.1188 → -0.6299),符合风控分箱预期。IV=0.0508,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 3 箱((1.5, inf])为极高风险区间:坏率高达 94.22%,策略上应重点管控

last6average — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 128.5]1493158133589.42%0.0278
2(128.5, 977.5]1289166112387.12%0.2502
3(977.5, 2255.5]1162113104990.28%-0.0663
4(2255.5, 3242.5]6498756286.59%0.2963
5(3242.5, inf]1956152180492.23%-0.3120
2026-11-03T11:57:45.242189 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(89.42% → 92.23% → 92.23%),如需严格单调可考虑调整切点或合并分箱。IV=0.0500,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 89.42% 回落至 87.12%,如需严格单调可考虑合并这两箱

第 5 箱((3242.5, inf])为极高风险区间:坏率高达 92.23%,策略上应重点管控

useratemore90monthsubcurmonth — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]8937681791.49%-0.2163
2(1.5, 3.5]7634771693.84%-0.5649
3(3.5, 5.5]7967272490.95%-0.1495
4(5.5, 8.5]97612085687.70%0.1938
5(8.5, inf]3430395303588.48%0.1195
2026-11-03T11:57:45.371044 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(91.49% → 93.84% → 88.48%),如需严格单调可考虑调整切点或合并分箱。IV=0.0497,该变量区分度偏弱但可用。

WOE 在第 3、4 箱之间穿越 0 线:前 3 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 91.49% 回升至 93.84%,如需严格单调可考虑合并这两箱

第 2 箱((1.5, 3.5])为极高风险区间:坏率高达 93.84%,策略上应重点管控

setl_int_avg — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 13.19]102712490387.93%0.1732
2(13.19, 16.74]3425328984.50%0.4625
3(16.74, 22.2]4723543792.58%-0.3660
4(22.2, 110.95]3287363292488.96%0.0723
5(110.95, inf]1730135159592.20%-0.3107
2026-11-03T11:57:45.498038 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.93% → 92.58% → 92.20%),如需严格单调可考虑调整切点或合并分箱。IV=0.0497,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 87.93% 回落至 84.50%,如需严格单调可考虑合并这两箱

第 3 箱((16.74, 22.2])为极高风险区间:坏率高达 92.58%,策略上应重点管控

bank_997_cnt_6m — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]3269391287888.04%0.1659
2(1.5, 3.5]1580129145191.84%-0.2582
3(3.5, 5.5]7838270189.53%0.0163
4(5.5, 9.5]5753753893.57%-0.5149
5(9.5, inf]3333629789.19%0.0518
2026-11-03T11:57:45.622068 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.04% → 93.57% → 89.19%),如需严格单调可考虑调整切点或合并分箱。IV=0.0484,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 91.84% 回落至 89.53%,如需严格单调可考虑合并这两箱

第 4 箱((5.5, 9.5])为极高风险区间:坏率高达 93.57%,策略上应重点管控

total_998_rate_3m — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -4999]5627648686.48%0.3066
2(-4999, 0.1]9459784889.74%-0.0061
3(0.1, 0.52]1997158183992.09%-0.2923
4(0.52, 0.73]1206115109190.46%-0.0879
5(0.73, inf]1830229160187.49%0.2174
2026-11-03T11:57:45.744890 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(86.48% → 92.09% → 87.49%),如需严格单调可考虑调整切点或合并分箱。IV=0.0481,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 92.09% 回落至 90.46%,如需严格单调可考虑合并这两箱

第 3 箱((0.1, 0.52])为极高风险区间:坏率高达 92.09%,策略上应重点管控

ovd_days_d15_num_m3 — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]2286177210992.26%-0.3192
2(0.5, 1.5]2769309246088.84%0.0840
3(1.5, 2.5]87810777187.81%0.1838
4(2.5, 4.5]5707949186.14%0.3316
5(4.5, inf]3553831789.30%0.0373
2026-11-03T11:57:45.875247 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(92.26% → 92.26% → 89.30%),如需严格单调可考虑调整切点或合并分箱。IV=0.0480,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 4→5 箱有局部反弹:Bad Rate 由 86.14% 回升至 89.30%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 0.5])为极高风险区间:坏率高达 92.26%,策略上应重点管控

loan_rate_1_12 — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.02]2511298221388.13%0.1570
2(0.02, 0.05]6236356089.89%-0.0228
3(0.05, 0.07]4402341794.77%-0.7355
4(0.07, 0.27]2371247212489.58%0.0104
5(0.27, inf]5954455192.61%-0.3655
2026-11-03T11:57:46.000101 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.13% → 94.77% → 92.61%),如需严格单调可考虑调整切点或合并分箱。IV=0.0479,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 94.77% 回落至 89.58%,如需严格单调可考虑合并这两箱

第 3 箱((0.05, 0.07])为极高风险区间:坏率高达 94.77%,策略上应重点管控

nosvaing_bank_rate_1_3 — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -4999]2050253179787.66%0.2016
2(-4999, 0.28]2580278230289.22%0.0481
3(0.28, 0.39]3392731292.04%-0.2851
4(0.39, 0.74]7995574493.12%-0.4427
5(0.74, inf]7726271091.97%-0.2761
2026-11-03T11:57:46.192399 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.66% → 93.12% → 91.97%),如需严格单调可考虑调整切点或合并分箱。IV=0.0466,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 4→5 箱有局部反弹:Bad Rate 由 93.12% 回落至 91.97%,如需严格单调可考虑合并这两箱

第 4 箱((0.39, 0.74])为极高风险区间:坏率高达 93.12%,策略上应重点管控

allcardsoverdraftrate6avg — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.07]3144368277688.30%0.1417
2(0.07, 0.88]1896154174291.88%-0.2634
3(0.88, 0.95]4285637286.92%0.2689
4(0.95, 0.99]3802535593.42%-0.4908
5(0.99, inf]6076354489.62%0.0066
2026-11-03T11:57:46.321527 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.30% → 93.42% → 89.62%),如需严格单调可考虑调整切点或合并分箱。IV=0.0457,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 91.88% 回落至 86.92%,如需严格单调可考虑合并这两箱

第 4 箱((0.95, 0.99])为极高风险区间:坏率高达 93.42%,策略上应重点管控

f_ps_od_days_max — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]5773548522590.51%-0.0963
2(0.5, 2.5]64210753583.33%0.5492
3(2.5, inf]4435538887.58%0.2049
2026-11-03T11:57:46.444901 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(90.51% → 90.51% → 87.58%),如需严格单调可考虑调整切点或合并分箱。IV=0.0454,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 83.33% 回升至 87.58%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 0.5])为极高风险区间:坏率高达 90.51%,策略上应重点管控

setl_rem_prcp_paym_m12_loan_os_prcp_pc — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -9998.5]4429425400490.40%-0.0843
2(-9998.5, -4999]5594151892.67%-0.3778
3(-4999, 17.22]80510570086.96%0.2615
4(17.22, 79.79]6369753984.75%0.4436
5(79.79, inf]4294238790.21%-0.0621
2026-11-03T11:57:46.567277 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(90.40% → 92.67% → 90.21%),如需严格单调可考虑调整切点或合并分箱。IV=0.0453,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 90.40% 回升至 92.67%,如需严格单调可考虑合并这两箱

第 2 箱((-9998.5, -4999])为极高风险区间:坏率高达 92.67%,策略上应重点管控

cardaveragemob — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 20.5]2142259188387.91%0.1786
2(20.5, 75.5]1833171166290.67%-0.1117
3(75.5, 84.5]4826242087.14%0.2493
4(84.5, 104.5]8775981893.27%-0.4669
5(104.5, inf]1121115100689.74%-0.0064
2026-11-03T11:57:46.695830 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.91% → 93.27% → 89.74%),如需严格单调可考虑调整切点或合并分箱。IV=0.0445,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 90.67% 回落至 87.14%,如需严格单调可考虑合并这两箱

第 4 箱((84.5, 104.5])为极高风险区间:坏率高达 93.27%,策略上应重点管控

sum_open_loan_mob24_neverdur — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]7109461686.76%0.2826
2(0.5, 3.5]3119357276288.55%0.1166
3(3.5, 4.5]7406967190.68%-0.1121
4(4.5, 8.5]1477117136092.08%-0.2905
5(8.5, inf]4102938192.93%-0.4129
2026-11-03T11:57:46.825080 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调上升(86.76% → 92.93%),WOE 单调下降(0.2826 → -0.4129),符合风控分箱预期。IV=0.0445,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 5 箱((8.5, inf])为极高风险区间:坏率高达 92.93%,策略上应重点管控

ncarloanbalancerate6month — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.19]1531181135088.18%0.1536
2(0.19, 0.63]3445129385.17%0.4147
3(0.63, 2.04]3446326312090.54%-0.0957
4(2.04, 2.81]3692034994.58%-0.6963
5(2.81, inf]7498666388.52%0.1206
2026-11-03T11:57:46.947255 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.18% → 94.58% → 88.52%),如需严格单调可考虑调整切点或合并分箱。IV=0.0444,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 88.18% 回落至 85.17%,如需严格单调可考虑合并这两箱

第 4 箱((2.04, 2.81])为极高风险区间:坏率高达 94.58%,策略上应重点管控

nosvaing_bank_rate_1m — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -4999]2424287213788.16%0.1544
2(-4999, 0.04]1864205165989.00%0.0711
3(0.04, 0.35]6937062389.90%-0.0240
4(0.35, 0.61]7606070092.11%-0.2947
5(0.61, inf]7995374693.37%-0.4824
2026-11-03T11:57:47.068758 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调上升(88.16% → 93.37%),WOE 单调下降(0.1544 → -0.4824),符合风控分箱预期。IV=0.0434,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 5 箱((0.61, inf])为极高风险区间:坏率高达 93.37%,策略上应重点管控

setl_prcp_avg — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 88.54]72010062086.11%0.3341
2(88.54, 609.66]3475375310089.21%0.0464
3(609.66, 719.26]3582833092.18%-0.3083
4(719.26, 936.76]4956343287.27%0.2333
5(936.76, inf]1810144166692.04%-0.2898
2026-11-03T11:57:47.194686 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(86.11% → 92.18% → 92.04%),如需严格单调可考虑调整切点或合并分箱。IV=0.0429,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 92.18% 回落至 87.27%,如需严格单调可考虑合并这两箱

第 3 箱((609.66, 719.26])为极高风险区间:坏率高达 92.18%,策略上应重点管控

bank_total_used_limit — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.16]112714598287.13%0.2492
2(0.16, 0.21]3062528191.83%-0.2574
3(0.21, 0.31]7648567988.87%0.0841
4(0.31, 0.75]3014257275791.47%-0.2108
5(0.75, inf]1329163116687.74%0.1945
2026-11-03T11:57:47.316556 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.13% → 91.83% → 87.74%),如需严格单调可考虑调整切点或合并分箱。IV=0.0426,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 91.83% 回落至 88.87%,如需严格单调可考虑合并这两箱

第 2 箱((0.16, 0.21])为极高风险区间:坏率高达 91.83%,策略上应重点管控

rh_cr_bal_pro — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -4999]3923435891.33%-0.1913
2(-4999, 0.45]112415097486.65%0.2921
3(0.45, 0.82]3011263274891.27%-0.1836
4(0.82, 0.97]1473157131689.34%0.0367
5(0.97, inf]4386037886.30%0.3223
2026-11-03T11:57:47.434108 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(91.33% → 91.33% → 86.30%),如需严格单调可考虑调整切点或合并分箱。IV=0.0418,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 86.65% 回升至 91.27%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, -4999])为极高风险区间:坏率高达 91.33%,策略上应重点管控

loan_reject_rate_6m — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.25]3002272273090.94%-0.1439
2(0.25, 0.51]1566181138588.44%0.1274
3(0.51, 0.62]4743444092.83%-0.3981
4(0.62, 0.69]4434739689.39%0.0311
5(0.69, inf]95013082086.32%0.3205
2026-11-03T11:57:47.630227 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(90.94% → 92.83% → 86.32%),如需严格单调可考虑调整切点或合并分箱。IV=0.0405,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 88.44% 回升至 92.83%,如需严格单调可考虑合并这两箱

第 3 箱((0.51, 0.62])为极高风险区间:坏率高达 92.83%,策略上应重点管控

no_bank_998_rate_12m — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.22]110195100691.37%-0.1978
2(0.22, 0.33]6514560693.09%-0.4382
3(0.33, 0.68]3833432340188.73%0.0986
4(0.68, 0.74]3455029585.51%0.3871
5(0.74, inf]6105355791.31%-0.1902
2026-11-03T11:57:47.756669 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(91.37% → 93.09% → 91.31%),如需严格单调可考虑调整切点或合并分箱。IV=0.0404,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 91.37% 回升至 93.09%,如需严格单调可考虑合并这两箱

第 2 箱((0.22, 0.33])为极高风险区间:坏率高达 93.09%,策略上应重点管控

max_setl_rem_prcp_paym_m6_last_month_loan_os_prcp_pc — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 3.39]3392383300988.71%0.0973
2(3.39, 9.1]5143947592.41%-0.3411
3(9.1, 11.68]5192949094.41%-0.6685
4(11.68, 90.9]1808199160988.99%0.0686
5(90.9, inf]6256056590.40%-0.0839
2026-11-03T11:57:47.883414 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.71% → 94.41% → 90.40%),如需严格单调可考虑调整切点或合并分箱。IV=0.0404,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 94.41% 回落至 88.99%,如需严格单调可考虑合并这两箱

第 3 箱((9.1, 11.68])为极高风险区间:坏率高达 94.41%,策略上应重点管控

c_bank_rate_6_12 — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.24]1353159119488.25%0.1459
2(0.24, 0.39]8428975389.43%0.0266
3(0.39, 0.44]3861936795.08%-0.7989
4(0.44, 0.68]2075194188190.65%-0.1097
5(0.68, inf]1884214167088.64%0.1074
2026-11-03T11:57:48.010370 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.25% → 95.08% → 88.64%),如需严格单调可考虑调整切点或合并分箱。IV=0.0394,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 95.08% 回落至 90.65%,如需严格单调可考虑合并这两箱

第 3 箱((0.39, 0.44])为极高风险区间:坏率高达 95.08%,策略上应重点管控

no_bank_total_used_limit — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.27]3482396308688.63%0.1088
2(0.27, 0.3]3802335793.95%-0.5802
3(0.3, 0.33]3754333288.53%0.1181
4(0.33, 0.58]1690140155091.72%-0.2423
5(0.58, inf]6137354088.09%0.1609
2026-11-03T11:57:48.137769 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.63% → 93.95% → 88.09%),如需严格单调可考虑调整切点或合并分箱。IV=0.0393,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 88.63% 回升至 93.95%,如需严格单调可考虑合并这两箱

第 2 箱((0.27, 0.3])为极高风险区间:坏率高达 93.95%,策略上应重点管控

mostcarduserate — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.15]3207369283888.49%0.1224
2(0.15, 0.82]6936562890.62%-0.1057
3(0.82, 0.99]7865373393.26%-0.4644
4(0.99, 1.07]1433134129990.65%-0.1091
5(1.07, inf]3364529186.61%0.2958
2026-11-03T11:57:48.269099 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.49% → 93.26% → 86.61%),如需严格单调可考虑调整切点或合并分箱。IV=0.0385,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 88.49% 回升至 90.62%,如需严格单调可考虑合并这两箱

第 3 箱((0.82, 0.99])为极高风险区间:坏率高达 93.26%,策略上应重点管控

cyc_accountcount — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 3.5]3314728485.80%0.3629
2(3.5, 10.5]3155350280588.91%0.0805
3(10.5, 11.5]4183338592.11%-0.2950
4(11.5, 16.5]1532165136789.23%0.0473
5(16.5, inf]110280102292.74%-0.3858
2026-11-03T11:57:48.401817 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(85.80% → 92.74% → 92.74%),如需严格单调可考虑调整切点或合并分箱。IV=0.0379,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 92.11% 回落至 89.23%,如需严格单调可考虑合并这两箱

第 5 箱((16.5, inf])为极高风险区间:坏率高达 92.74%,策略上应重点管控

total_new_cnt_rate_3m — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0]1372176119687.17%0.2458
2(0, 0.02]5694552492.09%-0.2928
3(0.02, 0.11]3195297289890.70%-0.1160
4(0.11, 0.17]7079361486.85%0.2747
5(0.17, inf]6976463390.82%-0.1295
2026-11-03T11:57:48.529758 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.17% → 92.09% → 90.82%),如需严格单调可考虑调整切点或合并分箱。IV=0.0377,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 92.09% 回落至 90.70%,如需严格单调可考虑合并这两箱

第 2 箱((0, 0.02])为极高风险区间:坏率高达 92.09%,策略上应重点管控

mini_loan_rate_6m — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.2]2483210227391.54%-0.2197
2(0.2, 0.37]2241260198188.40%0.1314
3(0.37, 0.48]7396867190.80%-0.1272
4(0.48, 0.57]5436248188.58%0.1133
5(0.57, inf]5347545985.96%0.3505
2026-11-03T11:57:48.651867 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(91.54% → 91.54% → 85.96%),如需严格单调可考虑调整切点或合并分箱。IV=0.0374,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 88.40% 回升至 90.80%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 0.2])为极高风险区间:坏率高达 91.54%,策略上应重点管控

apply_num_m12 — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]1549202134786.96%0.2612
2(0.5, 1.5]3111315279689.87%-0.0248
3(1.5, 2.5]103010592589.81%-0.0172
4(2.5, 3.5]4382940993.38%-0.4878
5(3.5, inf]7305967191.92%-0.2726
2026-11-03T11:57:48.778269 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(86.96% → 93.38% → 91.92%),如需严格单调可考虑调整切点或合并分箱。IV=0.0370,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 89.87% 回落至 89.81%,如需严格单调可考虑合并这两箱

第 4 箱((2.5, 3.5])为极高风险区间:坏率高达 93.38%,策略上应重点管控

loanearliestmonthsubcurmonth — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 47.5]7058262388.37%0.1346
2(47.5, 53.5]3605630484.44%0.4707
3(53.5, 85.5]2010209180189.60%0.0087
4(85.5, 93.5]6698458587.44%0.2216
5(93.5, inf]2711235247691.33%-0.1924
2026-11-03T11:57:48.907637 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.37% → 91.33% → 91.33%),如需严格单调可考虑调整切点或合并分箱。IV=0.0369,该变量区分度偏弱但可用。

WOE 在第 4、5 箱之间穿越 0 线:前 4 箱 WOE>0(客户质量优于整体),后 1 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 88.37% 回落至 84.44%,如需严格单调可考虑合并这两箱

第 5 箱((93.5, inf])为极高风险区间:坏率高达 91.33%,策略上应重点管控

loan_rate_3m_12m — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.07]71610361385.61%0.3787
2(0.07, 0.29]2145208193790.30%-0.0690
3(0.29, 0.34]6487956987.81%0.1879
4(0.34, 0.62]2052177187591.37%-0.1979
5(0.62, inf]8749777788.90%0.0816
2026-11-03T11:57:49.107063 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(85.61% → 91.37% → 88.90%),如需严格单调可考虑调整切点或合并分箱。IV=0.0364,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 90.30% 回落至 87.81%,如需严格单调可考虑合并这两箱

第 4 箱((0.34, 0.62])为极高风险区间:坏率高达 91.37%,策略上应重点管控

normal_repay_num — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 30.5]5008531447789.40%0.0267
2(30.5, 38.5]3812335893.96%-0.5864
3(38.5, 66.5]6917361889.44%0.0226
4(66.5, 103.5]3462232493.64%-0.5311
5(103.5, inf]4326137185.88%0.3533
2026-11-03T11:57:49.233713 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(89.40% → 93.96% → 85.88%),如需严格单调可考虑调整切点或合并分箱。IV=0.0363,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 89.40% 回升至 93.96%,如需严格单调可考虑合并这两箱

第 2 箱((30.5, 38.5])为极高风险区间:坏率高达 93.96%,策略上应重点管控

loanrecentlymonthsubcurmonth — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]2730230250091.58%-0.2235
2(1.5, 3.5]2334258207688.95%0.0772
3(3.5, 4.5]5346946587.08%0.2545
4(4.5, 6.5]5256446187.81%0.1879
5(6.5, inf]3324528786.45%0.3096
2026-11-03T11:57:49.356664 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(91.58% → 91.58% → 86.45%),如需严格单调可考虑调整切点或合并分箱。IV=0.0361,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 87.08% 回升至 87.81%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 1.5])为极高风险区间:坏率高达 91.58%,策略上应重点管控

setl_rem_prcp_paym_m3_num — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]4760540422088.66%0.1026
2(0.5, 1.5]1313119119490.94%-0.1473
3(1.5, inf]7855173493.50%-0.5081
2026-11-03T11:57:49.470662 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调上升(88.66% → 93.50%),WOE 单调下降(0.1026 → -0.5081),符合风控分箱预期。IV=0.0357,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 3 箱((1.5, inf])为极高风险区间:坏率高达 93.50%,策略上应重点管控

nosvaing_997_bank_cnt_6m — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]1184153103187.08%0.2542
2(0.5, 3.5]3469347312290.00%-0.0349
3(3.5, 4.5]5624651691.81%-0.2554
4(4.5, 6.5]6157653987.64%0.2031
5(6.5, inf]7105365792.54%-0.3553
2026-11-03T11:57:49.587120 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.08% → 92.54% → 92.54%),如需严格单调可考虑调整切点或合并分箱。IV=0.0347,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 91.81% 回落至 87.64%,如需严格单调可考虑合并这两箱

第 5 箱((6.5, inf])为极高风险区间:坏率高达 92.54%,策略上应重点管控

query_success_rate_6m — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.19]6509155986.00%0.3470
2(0.19, 0.41]1497173132488.44%0.1272
3(0.41, 0.69]2191210198190.42%-0.0819
4(0.69, 0.77]4285037888.32%0.1394
5(0.77, inf]1669140152991.61%-0.2284
2026-11-03T11:57:49.709459 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(86.00% → 91.61% → 91.61%),如需严格单调可考虑调整切点或合并分箱。IV=0.0338,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 90.42% 回落至 88.32%,如需严格单调可考虑合并这两箱

第 5 箱((0.77, inf])为极高风险区间:坏率高达 91.61%,策略上应重点管控

rh_ln_qy_fst_inte — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 675.5]2716298241889.03%0.0692
2(675.5, 688.5]4173138692.57%-0.3590
3(688.5, 706.5]94511583087.83%0.1863
4(706.5, 724.5]1517122139591.96%-0.2738
5(724.5, inf]8439874588.37%0.1344
2026-11-03T11:57:49.829384 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(89.03% → 92.57% → 88.37%),如需严格单调可考虑调整切点或合并分箱。IV=0.0331,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 89.03% 回升至 92.57%,如需严格单调可考虑合并这两箱

第 2 箱((675.5, 688.5])为极高风险区间:坏率高达 92.57%,策略上应重点管控

nosvaing_998_bank_rate_6m — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -3.75]95911584488.01%0.1688
2(-3.75, -1.13]4162639093.75%-0.5460
3(-1.13, 0.26]2327258206988.91%0.0802
4(0.26, 0.76]1797157164091.26%-0.1842
5(0.76, inf]104111992288.57%0.1146
2026-11-03T11:57:49.954815 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.01% → 93.75% → 88.57%),如需严格单调可考虑调整切点或合并分箱。IV=0.0330,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 93.75% 回落至 88.91%,如需严格单调可考虑合并这两箱

第 2 箱((-3.75, -1.13])为极高风险区间:坏率高达 93.75%,策略上应重点管控

total_997_cnt_1m — 4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]3662428323488.31%0.1397
2(0.5, 1.5]1632139149391.48%-0.2120
3(1.5, 2.5]6576459390.26%-0.0643
4(2.5, inf]5894454592.53%-0.3546
2026-11-03T11:57:50.071386 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.31% → 92.53% → 92.53%),如需严格单调可考虑调整切点或合并分箱。IV=0.0321,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 91.48% 回落至 90.26%,如需严格单调可考虑合并这两箱

第 4 箱((2.5, inf])为极高风险区间:坏率高达 92.53%,策略上应重点管控

no_bank_rate_3_6 — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.16]112814898086.88%0.2717
2(0.16, 0.3]7036364091.04%-0.1563
3(0.3, 0.36]4756441186.53%0.3023
4(0.36, 0.53]1418143127589.92%-0.0258
5(0.53, inf]2816257255990.87%-0.1363
2026-11-03T11:57:50.183593 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(86.88% → 91.04% → 90.87%),如需严格单调可考虑调整切点或合并分箱。IV=0.0318,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 91.04% 回落至 86.53%,如需严格单调可考虑合并这两箱

第 2 箱((0.16, 0.3])为极高风险区间:坏率高达 91.04%,策略上应重点管控

nosvaing_bank_org_3m — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]2050253179787.66%0.2016
2(0.5, 1.5]1769186158389.49%0.0207
3(1.5, 2.5]1134102103291.01%-0.1522
4(2.5, 4.5]109788100991.98%-0.2773
5(4.5, inf]4904644490.61%-0.1051
2026-11-03T11:57:50.303808 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.66% → 91.98% → 90.61%),如需严格单调可考虑调整切点或合并分箱。IV=0.0300,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 4→5 箱有局部反弹:Bad Rate 由 91.98% 回落至 90.61%,如需严格单调可考虑合并这两箱

第 4 箱((2.5, 4.5])为极高风险区间:坏率高达 91.98%,策略上应重点管控

c_bank_credit_used_limit — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.09]3194373282188.32%0.1388
2(0.09, 0.73]2129187194291.22%-0.1783
3(0.73, 0.83]3422531792.69%-0.3780
4(0.83, 0.99]4765542188.45%0.1267
5(0.99, inf]3993536491.23%-0.1798
2026-11-03T11:57:50.493104 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.32% → 92.69% → 91.23%),如需严格单调可考虑调整切点或合并分箱。IV=0.0291,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 92.69% 回落至 88.45%,如需严格单调可考虑合并这两箱

第 3 箱((0.73, 0.83])为极高风险区间:坏率高达 92.69%,策略上应重点管控

c_bank_rate_1_6 — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.12]4320472384889.07%0.0637
2(0.12, 0.23]5995254791.32%-0.1912
3(0.23, 0.33]3782135794.44%-0.6712
4(0.33, 0.59]8249273288.83%0.0881
5(0.59, inf]4193838190.93%-0.1432
2026-11-03T11:57:50.616412 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(89.07% → 94.44% → 90.93%),如需严格单调可考虑调整切点或合并分箱。IV=0.0281,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 94.44% 回落至 88.83%,如需严格单调可考虑合并这两箱

第 3 箱((0.23, 0.33])为极高风险区间:坏率高达 94.44%,策略上应重点管控

no_bank_cyc_used_limit — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.24]4080427365389.53%0.0155
2(0.24, 0.27]3304728385.76%0.3667
3(0.27, 0.33]5884953991.67%-0.2359
4(0.33, 0.4]4903345793.27%-0.4661
5(0.4, inf]105211993388.69%0.1028
2026-11-03T11:57:50.742564 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(89.53% → 93.27% → 88.69%),如需严格单调可考虑调整切点或合并分箱。IV=0.0278,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 85.76% 回升至 91.67%,如需严格单调可考虑合并这两箱

第 4 箱((0.33, 0.4])为极高风险区间:坏率高达 93.27%,策略上应重点管控

注:统计入模变量经合并调整后的最终分箱明细(开发样本口径);本次无变量发生单调性合并,细/粗分箱一致。

05 变量筛选

项目内容
筛选优先级黑名单 > 白名单 > 自动规则
IV 下限0.02
缺失率上限0.8
集中度上限0.9
相关性上限(皮尔逊)0.7
PSI 上限0.25
重要性 Top-K80
黑名单无
白名单无

候选变量 581,成功分箱 568,入模 80

变量白名单入模IV缺失率集中度IV门槛相关性PSI剔除原因
loan_no—N-0.0002PASSPASSFAIL——IV<0.02
last_con_amt—N0.0656PASSPASSPASS—0.3826PSI>0.25
sum_apply_amt—Y0.0571PASSPASSPASSPASS0.2029
max_apply_amt—N0.0308PASSPASSPASSFAIL0.1254与 sum_apply_amt 相关>0.7
min_apply_amt—N0.0175PASSPASSFAIL——IV<0.02
avg_apply_amt—N0.0306PASSPASSPASSFAIL0.1460与 apply_amt_avg_m12 相关>0.7
valid_sum_apply_amt—N0.0481PASSPASSPASSFAIL0.1739与 sum_apply_amt 相关>0.7
last_apply_amt—N0.0326PASSPASSPASSFAIL0.1318与 apply_amt_avg_m12 相关>0.7
f_apply_amt—N0.0340PASSPASSPASSFAIL0.0982与 sum_apply_amt 相关>0.7
apply_num—N0.0163PASSPASSFAIL——IV<0.02
max_apprv_amt—N0.0319PASSPASSPASS—1.4350PSI>0.25
min_apprv_amt—N0.0542PASSPASSPASS—1.3329PSI>0.25
avg_apprv_amt—N0.0287PASSPASSPASS—1.3944PSI>0.25
last_apprv_amt—N0.0352PASSPASSPASS—1.4248PSI>0.25
f_apprv_amt—N0.0397PASSPASSPASS—1.3992PSI>0.25
apprv_num—N0.0116PASSPASSFAIL——IV<0.02
sum_orig_prcp—Y0.0720PASSPASSPASSPASS0.2245
max_orig_prcp—N0.0800PASSPASSPASS—0.3803PSI>0.25
min_orig_prcp—N0.0371PASSPASSPASS—0.4179PSI>0.25
avg_orig_prcp—N0.0647PASSPASSPASS—0.4117PSI>0.25
last_orig_prcp—N0.0651PASSPASSPASS—0.3879PSI>0.25
f_orig_prcp—N0.0606PASSPASSPASS—0.4090PSI>0.25
loan_num—N0.0121PASSPASSFAIL——IV<0.02
loan_os_prcp—N0.1534PASSPASSPASS—0.3137PSI>0.25
pp_num—N0.0226PASSPASSPASSFAIL0.1118与 normal_repay_num 相关>0.7
apply_amt_day—N0.0000PASSFAIL———集中度>0.9
apprv_amt_day—N0.0000PASSFAIL———集中度>0.9
f_loan_ps_od_ind—N0.0055PASSPASSFAIL——IV<0.02
f_loan_orig_prcp—N0.0606PASSPASSPASS—0.4090PSI>0.25
rep_loan_num—N0.0121PASSPASSFAIL——IV<0.02
rep_loan_orig_prcp—N0.0324PASSPASSPASSFAIL0.1589与 sum_orig_prcp 相关>0.7
rep_loan_freq_avg—N0.0405PASSPASSPASSFAIL0.1792与 end_loan_to_cur_freq 相关>0.7
end_loan_to_rep_loan_freq—N0.0242PASSPASSPASSFAIL0.2141与 end_loan_to_cur_freq 相关>0.7
end_loan_to_cur_freq—Y0.1019PASSPASSPASSPASS0.2156
f_rep_loan_to_f_apply_dt_freq—N0.0262PASSPASSPASSPASS0.0958重要性排名超出Top-80
ps_prcp_amt_sum—N0.0352PASSPASSPASSFAIL0.0338与 sum_orig_prcp 相关>0.7
ps_int_amt_sum—N0.0241PASSPASS———唯一率过高(ID列)
ps_fee_amt_sum—N0.0000PASSFAIL———集中度>0.9
ps_prcp_amt_avg—N0.0586PASSPASSPASS—0.2583PSI>0.25
ps_int_amt_avg—N0.0765PASSPASS———唯一率过高(ID列)
ps_fee_amt_avg—N0.0000PASSFAIL———集中度>0.9
setl_prcp_mon—N0.0490PASSPASSPASSFAIL0.0465与 setl_prcp_mon_ratio 相关>0.7
setl_int_mon—Y0.0652PASSPASSPASSPASS0.0322
setl_fee_amt_mon—N0.0000PASSFAIL———集中度>0.9
setl_prcp_sum—N0.0308PASSPASSPASSFAIL0.0701与 sum_orig_prcp 相关>0.7
setl_int_sum—N0.0221PASSPASSPASSFAIL0.0714与 sum_orig_prcp 相关>0.7
setl_fee_amt_sum—N0.0000PASSFAIL———集中度>0.9
setl_prcp_avg—Y0.0429PASSPASSPASSPASS0.1761
setl_int_avg—Y0.0497PASSPASSPASSPASS0.2302
setl_fee_amt_avg—N0.0000PASSFAIL———集中度>0.9
setl_prcp_mon_ratio—Y0.0556PASSPASSPASSPASS0.0399
acc_ps_perd_no—Y0.2055PASSPASSPASSPASS0.1017
m_c_over_due_cnt—Y0.2452PASSPASSPASSPASS0.0162
cur_max_od_days—N0.0252PASSPASSPASSPASS0.0635重要性排名超出Top-80
max_od_days—N0.1186PASSPASSPASSFAIL0.0291与 max_od_days_m12 相关>0.7
m_s_ps_instm_amt—N0.0366PASSPASSPASSFAIL0.1871与 cur_od_loan_amt 相关>0.7
cur_od_loan_num—N0.0077PASSPASSFAIL——IV<0.02
cur_od_loan_amt—Y0.0936PASSPASSPASSPASS0.1478
f_ps_od_days_max—Y0.0454PASSPASSPASSPASS0.0471
s_ps_od_days_max—N0.0439PASSPASSPASSFAIL0.0105与 f_ps_od_days_max 相关>0.7
t_ps_od_days_max—Y0.0523PASSPASSPASSPASS0.0110
m1_ovd_days_num—N0.0755PASSFAIL———集中度>0.9
normal_repay_num—Y0.0363PASSPASSPASSPASS0.1011
numberresolve—N0.0002PASSFAIL———集中度>0.9
pboc_quiet_ind—N0.0000PASSFAIL———集中度>0.9
pboc_b_white_ind—N0.0000PASSFAIL———集中度>0.9
pboc_s_white_ind—N0.0000PASSFAIL———集中度>0.9
credit_ability—N-0.0002PASSFAIL———集中度>0.9
pboc_max—N0.0185PASSPASSFAIL——IV<0.02
pboc_min—N0.0002PASSFAIL———集中度>0.9
f_pboc—N0.0150PASSPASSFAIL——IV<0.02
last_pboc—N0.0002PASSFAIL———集中度>0.9
mobile_ovd_d30_idno_num—N0.0000PASSFAIL———集中度>0.9
ovd_days_max_m1—N0.0491PASSPASSPASSFAIL0.0531与 max_od_days_m12 相关>0.7
ovd_days_max_m3—N0.1060PASSPASSPASSFAIL0.0256与 max_od_days_m12 相关>0.7
ovd_days_max_m6—N0.1190PASSPASSPASSFAIL0.0280与 max_od_days_m12 相关>0.7
ovd_days_d3_num_m1—N0.0121PASSPASSFAIL——IV<0.02
ovd_days_d15_num_m3—Y0.0480PASSPASSPASSPASS0.0731
ovd_days_d30_num_m6—Y0.1315PASSPASSPASSPASS0.0296
last_ovd_to_last_applydt_freq—N0.0386PASSPASSPASS—0.7877PSI>0.25
last_ovd_to_last_setl_dt_freq—N0.0662PASSPASSPASS—7.4210PSI>0.25
login_num_d7—N0.0000PASSFAIL———集中度>0.9
login_num_d30—N0.0000PASSFAIL———集中度>0.9
login_num_d90—N0.0000PASSFAIL———集中度>0.9
balance_avg_m3—N0.1204PASSPASSPASS—0.3097PSI>0.25
balance_avg_m6—N0.1135PASSPASSPASS—0.2797PSI>0.25
balance_avg_m12—N0.0799PASSPASSPASS—0.2558PSI>0.25
rpym_fail_num_m3—N0.0000PASSFAIL———集中度>0.9
rpym_fail_num_m6—N0.0000PASSFAIL———集中度>0.9
rpym_fail_num_m12—N0.0000PASSFAIL———集中度>0.9
orig_prcp_avg_m1—N0.0000PASSFAIL———集中度>0.9
apply_amt_avg_m1—N0.0000PASSFAIL———集中度>0.9
orig_prcp_avg_m3—N0.0429PASSFAIL———集中度>0.9
apply_amt_avg_m3—N0.0420PASSFAIL———集中度>0.9
orig_prcp_avg_m6—Y0.0769PASSPASSPASSPASS0.0921
apply_amt_avg_m6—N0.0759PASSPASSPASSFAIL0.0570与 orig_prcp_avg_m6 相关>0.7
orig_prcp_avg_m12—N0.1218PASSPASSPASS—0.3549PSI>0.25
apply_amt_avg_m12—Y0.0793PASSPASSPASSPASS0.1669
suc_apply_trench_num—N0.0039PASSPASSFAIL——IV<0.02
suc_apply_top_pc_code_num—N0.0000PASSFAIL———集中度>0.9
suc_apply_sec_pc_code_num—N0.0000PASSFAIL———集中度>0.9
apply_num_d30—N0.0000PASSFAIL———集中度>0.9
apply_top_pc_code_num_d30—N0.0000PASSFAIL———集中度>0.9
apply_sec_pc_code_num_d30—N0.0000PASSFAIL———集中度>0.9
suc_apply_num_m1—N0.0000PASSFAIL———集中度>0.9
suc_apprv_num_m1—N0.0000PASSFAIL———集中度>0.9
apply_num_m1—N0.0000PASSFAIL———集中度>0.9
apprv_num_m1—N0.0000PASSFAIL———集中度>0.9
suc_apply_num_m3—N0.0260PASSFAIL———集中度>0.9
suc_apprv_num_m3—N0.0000PASSFAIL———集中度>0.9
apply_num_m3—N0.0258PASSPASSPASSPASS0.0634重要性排名超出Top-80
apprv_num_m3—N0.0000PASSFAIL———集中度>0.9
suc_apply_num_m6—N0.0485PASSPASSPASSFAIL0.0380与 apply_num_m6 相关>0.7
suc_apprv_num_m6—N0.0049PASSPASSFAIL——IV<0.02
apply_num_m6—Y0.0555PASSPASSPASSPASS0.0347
apprv_num_m6—N0.0028PASSPASSFAIL——IV<0.02
suc_apply_num_m12—N0.0358PASSPASSPASSFAIL0.1874与 apply_num_m12 相关>0.7
suc_apprv_num_m12—N0.0162PASSPASSFAIL——IV<0.02
apply_num_m12—Y0.0370PASSPASSPASSPASS0.2194
apprv_num_m12—N0.0126PASSPASSFAIL——IV<0.02
last_apply_od_num_m12—N0.0062PASSPASSFAIL——IV<0.02
last_apply_od_num_m1—N0.0070PASSPASSFAIL——IV<0.02
last_apply_od_num_m3—N0.0067PASSPASSFAIL——IV<0.02
last_apply_od_num_m6—N0.0062PASSPASSFAIL——IV<0.02
apply_trench_num—N0.0021PASSPASSFAIL——IV<0.02
night_apply_num—N—FAILPASS———分箱失败(单侧样本)
last_apply_freq—N0.0211PASSPASSPASSPASS0.0246重要性排名超出Top-80
red_envelope_num—N0.0000PASSFAIL———集中度>0.9
edu_loan_num—N0.0000PASSFAIL———集中度>0.9
tra_loan_num—N0.0000PASSFAIL———集中度>0.9
oth_loan_num—N0.0000PASSFAIL———集中度>0.9
gex_loan_num—N0.0135PASSPASSFAIL——IV<0.02
cash_loan_num—N0.0136PASSPASSFAIL——IV<0.02
med_loan_num—N0.0000PASSFAIL———集中度>0.9
trans_loan_num—N0.0000PASSFAIL———集中度>0.9
dec_loan_num—N0.0000PASSFAIL———集中度>0.9
ren_loan_num—N0.0000PASSFAIL———集中度>0.9
court_black_list—N—FAILPASS———分箱失败(单侧样本)
red_list—N—FAILPASS———分箱失败(单侧样本)
pengyuan_black_list—N—FAILPASS———分箱失败(单侧样本)
samoye_black_list—N—FAILPASS———分箱失败(单侧样本)
breach_of_person—N—FAILPASS———分箱失败(单侧样本)
tongdun_black_list—N—FAILPASS———分箱失败(单侧样本)
bh_cur_bad_num—N—FAILPASS———分箱失败(单侧样本)
bh_cur_od_num—N0.0163PASSPASSFAIL——IV<0.02
hj_cur_bad_num—N0.0055PASSPASSFAIL——IV<0.02
hj_cur_od_num—N0.0055PASSPASSFAIL——IV<0.02
sub3_mobile—N0.0275PASSPASSPASSPASS0.0070重要性排名超出Top-80
mobile_idno_num—N—FAILPASS———分箱失败(单侧样本)
work_phone_idno_num—N—FAILPASS———分箱失败(单侧样本)
mobile_idno_num_m1—N0.0000PASSFAIL———集中度>0.9
mobile_idno_num_m3—N0.0256PASSPASSPASSFAIL0.0636与 apply_num_m3 相关>0.7
mobile_idno_num_m6—N0.0530PASSPASSPASSFAIL0.0309与 repay_amount_m6_loan_os_prcp_pc 相关>0.7
mobile_idno_num_m12—N0.0228PASSPASSPASSPASS0.1885重要性排名超出Top-80
setl_prcp_ratio—N0.1266PASSPASSPASSFAIL0.1736与 loan_os_prcp_orig_prcp_pc 相关>0.7
last_loan_setl_tnr—N0.0978PASSPASSPASSFAIL0.1032与 loan_os_prcp_orig_prcp_pc 相关>0.7
last_od_num_m3—N0.1321PASSPASSPASSFAIL0.1340与 acc_ps_perd_no 相关>0.7
last_od_num_m6—N0.1688PASSPASSPASSFAIL0.1228与 acc_ps_perd_no 相关>0.7
last_od_num_m12—N0.1785PASSPASSPASSFAIL0.1180与 acc_ps_perd_no 相关>0.7
last_od_to_cur_freq—N0.0000PASSFAIL———集中度>0.9
last_clear_num_m1—N0.0000PASSFAIL———集中度>0.9
last_clear_num_m3—N0.0108PASSPASSFAIL——IV<0.02
last_clear_num_m6—N0.0034PASSPASSFAIL——IV<0.02
last_clear_num_m12—N0.0071PASSPASSFAIL——IV<0.02
ovd_days_max_m12—N0.1204PASSPASSPASSFAIL0.0283与 max_od_days_m12 相关>0.7
ovd_days_d3_count_m3—N0.1350PASSPASSPASSFAIL0.1124与 acc_ps_perd_no 相关>0.7
ovd_days_d3_count_m6—N0.1752PASSPASSPASSFAIL0.1086与 acc_ps_perd_no 相关>0.7
ovd_days_d3_count_m12—N0.1807PASSPASSPASSFAIL0.1067与 acc_ps_perd_no 相关>0.7
ovd_days_d30_count_m3—N0.0659PASSFAIL———集中度>0.9
ovd_days_d30_count_m6—N0.0754PASSFAIL———集中度>0.9
ovd_days_d30_count_m12—N0.0772PASSFAIL———集中度>0.9
ovd_days_d3_month_m3—N0.3097PASSPASSPASSFAIL0.0273与 odgt0_m12_num 相关>0.7
ovd_days_d3_month_m6—N0.3198PASSPASSPASSFAIL0.0395与 odgt0_m12_num 相关>0.7
ovd_days_d3_month_m12—N0.3198PASSPASSPASSFAIL0.0395与 odgt0_m12_num 相关>0.7
ovd_days_d30_month_m3—N0.0755PASSFAIL———集中度>0.9
ovd_days_d30_month_m6—N0.0755PASSFAIL———集中度>0.9
ovd_days_d30_month_m12—N0.0755PASSFAIL———集中度>0.9
loan_os_prcp_claim—N0.1534PASSPASSPASS—0.3137PSI>0.25
balance_avg_m3_claim—N0.1204PASSPASSPASS—0.3097PSI>0.25
balance_avg_m6_claim—N0.1135PASSPASSPASS—0.2797PSI>0.25
balance_avg_m12_claim—N0.0799PASSPASSPASS—0.2558PSI>0.25
account_age—Y0.0555PASSPASSPASSPASS0.0525
account_age_loan_day—Y0.0736PASSPASSPASSPASS0.1271
max_od_due—N0.0718PASSFAIL———集中度>0.9
odgt0_m3_mob_pc—N0.2926PASSPASSPASSFAIL0.0162与 odgt0_m12_num 相关>0.7
odgt0_m6_mob_pc—N0.3159PASSPASSPASSFAIL0.0716与 odgt0_m12_num 相关>0.7
odgt0_m12_mob_pc—Y0.2573PASSPASSPASSPASS0.0678
odgt0_m3_num—N0.3063PASSPASSPASSFAIL0.0479与 odgt0_m12_num 相关>0.7
odgt0_m6_num—N0.3408PASSPASSPASSFAIL0.0604与 odgt0_m12_num 相关>0.7
odgt0_m12_num—Y0.3489PASSPASSPASSPASS0.0597
odgt1_m3_mob_pc—N0.0594PASSFAIL———集中度>0.9
odgt1_m6_mob_pc—N0.0720PASSFAIL———集中度>0.9
odgt1_m12_mob_pc—N0.0738PASSFAIL———集中度>0.9
odgt1_m3_num—N0.0594PASSFAIL———集中度>0.9
odgt1_m6_num—N0.0720PASSFAIL———集中度>0.9
odgt1_m12_num—N0.0738PASSFAIL———集中度>0.9
odgt2_m3_mob_pc—N0.0000PASSFAIL———集中度>0.9
odgt2_m6_mob_pc—N0.0000PASSFAIL———集中度>0.9
odgt2_m12_mob_pc—N0.0000PASSFAIL———集中度>0.9
odgt2_m3_num—N0.0000PASSFAIL———集中度>0.9
odgt2_m6_num—N0.0000PASSFAIL———集中度>0.9
odgt2_m12_num—N0.0000PASSFAIL———集中度>0.9
odgt3_m3_mob_pc—N0.0000PASSFAIL———集中度>0.9
odgt3_m6_mob_pc—N0.0000PASSFAIL———集中度>0.9
odgt3_m12_mob_pc—N0.0000PASSFAIL———集中度>0.9
odgt3_m3_num—N0.0000PASSFAIL———集中度>0.9
odgt3_m6_num—N0.0000PASSFAIL———集中度>0.9
odgt3_m12_num—N0.0000PASSFAIL———集中度>0.9
odgt0_m12_subcurmonth—N0.0000PASSFAIL———集中度>0.9
odgt1_m12_subcurmonth—N0.0738PASSFAIL———集中度>0.9
odgt2_m12_subcurmonth—N0.0000PASSFAIL———集中度>0.9
odgt3_m12_subcurmonth—N0.0000PASSFAIL———集中度>0.9
max_od_due_m3—N0.0597PASSFAIL———集中度>0.9
max_od_due_m6—N0.0720PASSFAIL———集中度>0.9
max_od_due_m12—N0.0738PASSFAIL———集中度>0.9
max_od_days_m3—N0.1068PASSPASSPASSFAIL0.0251与 max_od_days_m12 相关>0.7
max_od_days_m6—N0.1199PASSPASSPASSFAIL0.0275与 max_od_days_m12 相关>0.7
max_od_days_m12—Y0.1213PASSPASSPASSPASS0.0278
odgt0_conti_max_mons_m12—N0.3367PASSPASSPASSFAIL0.0499与 odgt0_m12_num 相关>0.7
odgt1_conti_max_mons_m12—N0.0738PASSFAIL———集中度>0.9
odgt2_conti_max_mons_m12—N0.0000PASSFAIL———集中度>0.9
odgt3_conti_max_mons_m12—N0.0000PASSFAIL———集中度>0.9
tothismonthodcontiaddnumm12—Y0.1837PASSPASSPASSPASS0.0301
tothismonthodcontireducenumm12—N0.0000PASSFAIL———集中度>0.9
odduecontiaddnumm12—N0.0348PASSFAIL———集中度>0.9
odduecontireducenumm12—Y0.0908PASSPASSPASSPASS0.0040
odmaxreduceduem12—N0.0908PASSPASSPASSFAIL0.0040与 odduecontireducenumm12 相关>0.7
loan_os_prcp_orig_prcp_pc—Y0.1437PASSPASSPASSPASS0.1869
od_amt_loan_os_prcp_pc—N0.1174PASSPASSPASS—0.5406PSI>0.25
od_amt_m3_loan_os_prcp_m3_pc—Y0.1301PASSPASSPASSPASS0.0184
od_amt_m6_loan_os_prcp_m6_pc—N0.1159PASSPASSPASS—0.3258PSI>0.25
od_amt_m12_loan_os_prcp_m12_pc—Y0.0654PASSPASSPASSPASS0.2101
od_amt_orig_prcp_pc—N0.0610PASSPASSPASS—0.6222PSI>0.25
avg_od_amt_m3_orig_prcp_pc—N0.0878PASSPASSPASS—0.4432PSI>0.25
avg_od_amt_m6_orig_prcp_pc—N0.0751PASSPASSPASS—0.5300PSI>0.25
avg_od_amt_m12_orig_prcp_pc—N0.0441PASSPASSPASS—0.3028PSI>0.25
ps_od_norm_int_amt—Y0.1186PASSPASSPASSPASS0.2487
max_loan_os_prcp_m3—N0.1181PASSPASSPASS—0.3345PSI>0.25
max_loan_os_prcp_m6—N0.0965PASSPASSPASS—0.2701PSI>0.25
max_loan_os_prcp_m12—N0.0642PASSPASSPASSFAIL0.2439与 sum_orig_prcp 相关>0.7
loan_os_prcp_gt0_m3_num—N0.0000PASSFAIL———集中度>0.9
loan_os_prcp_gt0_m6_num—N0.0251PASSPASSPASSFAIL0.0655与 setl_rem_prcp_paym_m3_loan_os_prcp_pc 相关>0.7
loan_os_prcp_gt0_m12_num—N0.0461PASSPASSPASSFAIL0.0761与 repay_amount_m6_loan_os_prcp_pc 相关>0.7
loan_os_prcp_gt0_m3_num_pc—N0.0000PASSFAIL———集中度>0.9
loan_os_prcp_gt0_m6_num_pc—N0.0010PASSFAIL———集中度>0.9
loan_os_prcp_gt0_m12_num_pc—N0.0186PASSPASSFAIL——IV<0.02
loanosprcpgt0subcurmonth—N0.0000PASSFAIL———集中度>0.9
loanosprcpgt0m12maxcontimons—N0.0454PASSPASSPASSFAIL0.0670与 repay_amount_m6_loan_os_prcp_pc 相关>0.7
repay_amount_last_month_loan_os_prcp_pc—Y0.1273PASSPASSPASSPASS0.1697
repay_amount_m3_loan_os_prcp_pc—N0.1131PASSPASSPASS—0.2789PSI>0.25
repay_amount_m6_loan_os_prcp_pc—Y0.0729PASSPASSPASSPASS0.2336
repay_amount_m12_loan_os_prcp_pc—N0.0318PASSPASSPASSFAIL0.0507与 setl_rem_prcp_paym_m12_loan_os_prcp_pc 相关>0.7
all_repay_amount_shd_amt_pc—N0.0980PASSPASSPASSFAIL0.1052与 repay_amount_all_shd_amt_pc 相关>0.7
all_repay_amount_m3_shd_amt_pc—Y0.1342PASSPASSPASSPASS0.1558
all_repay_amount_m6_shd_amt_pc—N0.0875PASSPASSPASSFAIL0.1764与 all_repay_amount_m3_shd_amt_pc 相关>0.7
all_repay_amount_m12_shd_amt_pc—N0.0849PASSPASSPASSFAIL0.1476与 all_repay_amount_m3_shd_amt_pc 相关>0.7
repay_amount_all_shd_amt_pc—Y0.1084PASSPASSPASSPASS0.1174
repay_amount_m3_all_shd_amt_pc—N0.0651PASSPASSPASS—0.3780PSI>0.25
repay_amount_m6_all_shd_amt_pc—N0.0497PASSPASSPASS—0.2841PSI>0.25
repay_amount_m12_all_shd_amt_pc—Y0.0527PASSPASSPASSPASS0.1581
setl_rem_prcp_paym_last_month_loan_os_prcp_pc—N0.0013PASSFAIL———集中度>0.9
setl_rem_prcp_paym_m3_loan_os_prcp_pc—Y0.0649PASSPASSPASSPASS0.1715
setl_rem_prcp_paym_m6_loan_os_prcp_pc—N0.0432PASSPASSPASSFAIL0.1261与 repay_amount_m6_loan_os_prcp_pc 相关>0.7
setl_rem_prcp_paym_m12_loan_os_prcp_pc—Y0.0453PASSPASSPASSPASS0.0344
max_setl_rem_prcp_paym_m3_last_month_loan_os_prcp_pc—Y0.0888PASSPASSPASSPASS0.1148
max_setl_rem_prcp_paym_m6_last_month_loan_os_prcp_pc—Y0.0404PASSPASSPASSPASS0.1923
max_setl_rem_prcp_paym_m12_last_month_loan_os_prcp_pc—N0.0129PASSPASSFAIL——IV<0.02
setl_rem_prcp_paym_m3_num—Y0.0357PASSPASSPASSPASS0.0322
setl_rem_prcp_paym_m6_num—N0.0293PASSPASSPASSFAIL0.0748与 setl_rem_prcp_paym_m3_num 相关>0.7
setl_rem_prcp_paym_m12_num—N0.0037PASSPASSFAIL——IV<0.02
loan_os_prcp_credit_amount—N0.0907PASSPASSPASS—0.6048PSI>0.25
loan_os_prcp_credit_amount_m3—N0.0716PASSPASSPASS—0.6533PSI>0.25
loan_os_prcp_credit_amount_m6—N0.0834PASSPASSPASS—0.7686PSI>0.25
loan_os_prcp_credit_amount_m12—N0.0638PASSPASSPASS—0.6855PSI>0.25
useratemore50_m3_num—N0.0511PASSPASSPASSFAIL0.1620与 loan_os_prcp_orig_prcp_pc 相关>0.7
useratemore50_m6_num—N0.0275PASSPASSPASS—0.3968PSI>0.25
useratemore50_m12_num—N0.0188PASSPASSFAIL——IV<0.02
useratemore50_m3_num_pc—N0.0511PASSPASSPASSFAIL0.1629与 loan_os_prcp_orig_prcp_pc 相关>0.7
useratemore50_m6_num_pc—N0.0481PASSPASSPASSFAIL0.2056与 useratemore90monthsubcurmonth 相关>0.7
useratemore50_m12_num_pc—N0.0445PASSPASSPASS—0.6259PSI>0.25
useratemore50monthsubcurmonth—N0.0575PASSPASSPASS—0.5423PSI>0.25
useratemore50maxcontimonsm12—N0.0178PASSPASSFAIL——IV<0.02
useratemore70_m3_num—N0.0448PASSPASSPASSFAIL0.1751与 useratemore90_m6_num 相关>0.7
useratemore70_m6_num—N0.0482PASSPASSPASSFAIL0.1940与 useratemore90_m6_num 相关>0.7
useratemore70_m12_num—N0.0139PASSPASSFAIL——IV<0.02
useratemore70_m3_num_pc—N0.0467PASSPASSPASSFAIL0.1793与 useratemore90_m6_num 相关>0.7
useratemore70_m6_num_pc—N0.0566PASSPASSPASSFAIL0.1790与 loan_os_prcp_orig_prcp_pc 相关>0.7
useratemore70_m12_num_pc—N0.0463PASSPASSPASSFAIL0.1994与 useratemore90monthsubcurmonth 相关>0.7
useratemore70monthsubcurmonth—N0.0571PASSPASSPASS—0.4424PSI>0.25
useratemore70maxcontimonsm12—N0.0148PASSPASSFAIL——IV<0.02
useratemore90_m3_num—N0.0144PASSPASSFAIL——IV<0.02
useratemore90_m6_num—Y0.0556PASSPASSPASSPASS0.0585
useratemore90_m12_num—N0.0320PASSPASSPASS—0.7297PSI>0.25
useratemore90_m3_num_pc—N0.0152PASSPASSFAIL——IV<0.02
useratemore90_m6_num_pc—N0.0548PASSPASSPASSFAIL0.0523与 useratemore90_m6_num 相关>0.7
useratemore90_m12_num_pc—N0.0446PASSPASSPASS—0.4660PSI>0.25
useratemore90monthsubcurmonth—Y0.0497PASSPASSPASSPASS0.1954
useratemore90maxcontimonsm12—N0.0292PASSPASSPASS—0.7186PSI>0.25
max_loan_os_prcp_m1—N0.1465PASSPASSPASS—0.3175PSI>0.25
orgsum1—N0.0279PASSPASSPASSFAIL0.0627与 nosvaing_bank_cnt_1m 相关>0.7
orgsum2—N0.0000PASSFAIL———集中度>0.9
recordsum1—N0.0316PASSPASSPASSFAIL0.0670与 nosvaing_bank_cnt_1m 相关>0.7
recordsum2—N0.0000PASSFAIL———集中度>0.9
recordsum3—N0.0042PASSFAIL———集中度>0.9
cardsum—N0.0345PASSPASSPASSFAIL0.0761与 sum_card_mob12_neverdur 相关>0.7
loanlaworg—N0.0239PASSPASSPASSPASS0.1436重要性排名超出Top-80
loanorg—N0.0000PASSFAIL———集中度>0.9
loancount—N0.0239PASSPASSPASSFAIL0.0382与 loanlaworg 相关>0.7
last6average—Y0.0500PASSPASSPASSPASS0.0458
cardlaworg—N0.0197PASSPASSFAIL——IV<0.02
cardorg—N0.0000PASSFAIL———集中度>0.9
cardaccountcount—N0.0396PASSPASSPASSFAIL0.0187与 sum_card_mob12_neverdur 相关>0.7
cyc_orgcount—N0.0282PASSPASSPASSFAIL0.0046与 cyc_accountcount 相关>0.7
cyc_accountcount—Y0.0379PASSPASSPASSPASS0.0049
mostcarduserate—Y0.0385PASSPASSPASSPASS0.1039
allcarduserate—N0.0000PASSFAIL———集中度>0.9
creditcardduetimes12m—N0.0044PASSFAIL———集中度>0.9
cardnumbers—N0.0336PASSPASSPASSFAIL0.0839与 sum_card_mob12_neverdur 相关>0.7
houseloannum—N0.0029PASSPASSFAIL——IV<0.02
loannum—Y0.0712PASSPASSPASSPASS0.0509
no_query_6m_cca—N0.0038PASSPASSFAIL——IV<0.02
p_mob3_neverdelq_total—N0.0971PASSPASSPASS—0.2649PSI>0.25
loanearliestmonthsubcurmonth—Y0.0369PASSPASSPASSPASS0.2133
cardearliestmonthsubcurmonth—N0.0334PASSPASSPASSFAIL0.1678与 cardaveragemob 相关>0.7
normaldjcardnumber—N0.0286PASSPASSPASSFAIL0.0814与 sum_card_mob12_neverdur 相关>0.7
creditquerytimes—N0.0132PASSPASSFAIL——IV<0.02
cardquerytimes—N0.0090PASSPASSFAIL——IV<0.02
loanquerytimes—N0.0196PASSPASSFAIL——IV<0.02
cardquerytimescur3month—N0.0036PASSPASSFAIL——IV<0.02
loanquerytimescur3month—N0.0229PASSPASSPASSFAIL0.0252与 bank_cnt_3m 相关>0.7
loanquery4in1monstring—N0.0304PASSPASSPASSFAIL0.0661与 nosvaing_bank_cnt_1m 相关>0.7
djcardquery4in1monstring—N0.0000PASSFAIL———集中度>0.9
allcardsoverdraftrate6avg—Y0.0457PASSPASSPASSPASS0.1372
sumcarduseratemore50—N0.0168PASSPASSFAIL——IV<0.02
sum_pcard_mob12_neverdur—N0.0000PASSFAIL———集中度>0.9
sum_card_mob12_neverdur—Y0.0657PASSPASSPASSPASS0.0898
sum_card_month6_m1—N0.0000PASSFAIL———集中度>0.9
sum_pcard_month6_m1—N0.0000PASSFAIL———集中度>0.9
sum_querytimes_month6—N0.0137PASSPASSFAIL——IV<0.02
cardquerytimescur12month—N0.0032PASSPASSFAIL——IV<0.02
sum_loan_month24_m2—N0.0000PASSFAIL———集中度>0.9
sum_card_month24_m2—N0.0000PASSFAIL———集中度>0.9
loanquerytimescur6month—N0.0099PASSPASSFAIL——IV<0.02
sum_card_min_pay—N0.0097PASSPASSFAIL——IV<0.02
sum_month3_new_loan—Y0.0591PASSPASSPASSPASS0.0811
sum_uncleared_loan—N0.0232PASSPASSPASSFAIL0.1120与 cyc_accountcount 相关>0.7
loanrecentlymonthsubcurmonth—Y0.0361PASSPASSPASSPASS0.1078
houseloannumst—N0.0027PASSPASSFAIL——IV<0.02
sum_loan_mob24_neverdur—N0.0705PASSPASSPASSFAIL0.1152与 loannum 相关>0.7
sum_card_mob24_neverdur—N0.0408PASSPASSPASSFAIL0.0913与 sum_card_mob12_neverdur 相关>0.7
sum_querytimes_month6st—N0.0129PASSPASSFAIL——IV<0.02
cardaveragemob—Y0.0445PASSPASSPASSPASS0.1555
sumcarduseratemore80—N0.0197PASSPASSFAIL——IV<0.02
loanquerytimescur2month—N0.0139PASSPASSFAIL——IV<0.02
cardquerytimescur2month—N0.0022PASSFAIL———集中度>0.9
sum_open_loan_mob24_neverdur—Y0.0445PASSPASSPASSPASS0.1749
sum_open_card_mob24_neverdur—N0.0348PASSPASSPASSFAIL0.0615与 sum_card_mob12_neverdur 相关>0.7
sum_open_pcard_mob24_neverdur—N0.0000PASSFAIL———集中度>0.9
cardrecentlymonthsubcurmonth—N0.0414PASSPASSPASSFAIL0.0565与 cardaveragemob 相关>0.7
loan_m12_query_num—N—FAILPASS———分箱失败(单侧样本)
card_m6_query_num—N—FAILPASS———分箱失败(单侧样本)
card_m12_query_num—N—FAILPASS———分箱失败(单侧样本)
rh_consln_clear_num_m12—N0.0059PASSPASSFAIL——IV<0.02
m1orgloan—N0.0277PASSPASSPASSFAIL0.0612与 nosvaing_bank_cnt_1m 相关>0.7
grxfloanclearednum—N0.0534PASSPASSPASSFAIL0.0639与 loannum 相关>0.7
ncarloanbalancerate6month—Y0.0444PASSPASSPASSPASS0.0213
mean_numl_rrp—N0.0128PASSPASSFAIL——IV<0.02
rh_cr_bal_pro—Y0.0418PASSPASSPASSPASS0.0065
rh_crc_uc_max_age—N0.0406PASSPASSPASSFAIL0.1521与 cardaveragemob 相关>0.7
rh_ln_qy_fst_inte—Y0.0331PASSPASSPASSPASS0.0118
rh_ln_y2_qyf_org_pro—N0.0066PASSFAIL———集中度>0.9
loan24mquerycnt—N0.0000PASSFAIL———集中度>0.9
m1orgcard—N0.0000PASSFAIL———集中度>0.9
loanquerytimescur2month_no_by—N0.0149PASSPASSFAIL——IV<0.02
no_bank_org_3m—N0.0217PASSPASSPASSFAIL0.0215与 nosvaing_bank_org_3m 相关>0.7
no_bank_org_1m—N0.0222PASSPASSPASSFAIL0.0862与 nosvaing_bank_cnt_1m 相关>0.7
no_bank_org_12m—N0.0150PASSPASSFAIL——IV<0.02
no_bank_reject_rate_6m—N0.0101PASSPASSFAIL——IV<0.02
loan_reject_rate_6m—Y0.0405PASSPASSPASSPASS0.0550
loan_reject_rate_3m—Y0.0569PASSPASSPASSPASS0.0683
loan_rate_3m_12m—Y0.0364PASSPASSPASSPASS0.1493
loan_org_1m—N0.0252PASSPASSPASSFAIL0.0697与 nosvaing_bank_cnt_1m 相关>0.7
loan_reject_rate_1m—N0.0243PASSPASSPASSPASS0.0318重要性排名超出Top-80
no_bank_org_6m—N0.0066PASSPASSFAIL——IV<0.02
query_success_rate_6m—Y0.0338PASSPASSPASSPASS0.0489
no_org_12m_cca—N0.0071PASSPASSFAIL——IV<0.02
loan_pass_rate_6m—N0.0299PASSPASSPASSFAIL0.0494与 loan_reject_rate_6m 相关>0.7
cardqueryatotalorg6month—N0.0047PASSPASSFAIL——IV<0.02
no_query_6m—N0.0109PASSPASSFAIL——IV<0.02
loanqueryatotalorg6month—N0.0124PASSPASSFAIL——IV<0.02
loanqueryatotalorg3month—N0.0235PASSPASSPASSFAIL0.0549与 bank_cnt_3m 相关>0.7
cardqueryatotalorg3month—N0.0031PASSPASSFAIL——IV<0.02
loan_cnt_12m—N0.0205PASSPASSPASSFAIL0.1969与 loanlaworg 相关>0.7
credithistorynum—N0.0436PASSPASSPASSFAIL0.2252与 cardaveragemob 相关>0.7
c_bank_org_1m—N0.0087PASSPASSFAIL——IV<0.02
c_bank_org_3m—N0.0166PASSPASSFAIL——IV<0.02
c_bank_org_6m—N0.0071PASSPASSFAIL——IV<0.02
c_bank_org_12m—N0.0200PASSPASSPASSFAIL0.1453与 loanlaworg 相关>0.7
bank_org_1m—N0.0096PASSPASSFAIL——IV<0.02
bank_org_3m—N0.0233PASSPASSPASSFAIL0.0512与 bank_cnt_3m 相关>0.7
bank_org_6m—N0.0081PASSPASSFAIL——IV<0.02
bank_org_12m—N0.0116PASSPASSFAIL——IV<0.02
total_org_1m—N0.0264PASSPASSPASSFAIL0.0717与 nosvaing_bank_cnt_1m 相关>0.7
total_org_3m—N0.0265PASSPASSPASSFAIL0.0687与 nosvaing_bank_org_3m 相关>0.7
total_org_6m—N0.0091PASSPASSFAIL——IV<0.02
total_org_12m—N0.0092PASSPASSFAIL——IV<0.02
nosvaing_bank_org_1m—N0.0481PASSPASSPASSFAIL0.1889与 nosvaing_bank_cnt_1m 相关>0.7
nosvaing_bank_org_3m—Y0.0300PASSPASSPASSPASS0.1033
nosvaing_bank_org_6m—N0.0107PASSPASSFAIL——IV<0.02
nosvaing_bank_org_12m—N0.0160PASSPASSFAIL——IV<0.02
mini_loan_org_1m—N0.0003PASSPASSFAIL——IV<0.02
mini_loan_org_3m—N0.0035PASSPASSFAIL——IV<0.02
mini_loan_org_6m—N0.0079PASSPASSFAIL——IV<0.02
mini_loan_org_12m—N0.0323PASSPASSPASS—0.5131PSI>0.25
no_bank_cnt_1m—N0.0215PASSPASSPASSFAIL0.0968与 nosvaing_bank_cnt_1m 相关>0.7
no_bank_cnt_3m—N0.0214PASSPASSPASSFAIL0.0252与 nosvaing_bank_org_3m 相关>0.7
no_bank_cnt_6m—N0.0078PASSPASSFAIL——IV<0.02
no_bank_cnt_12m—N0.0158PASSPASSFAIL——IV<0.02
bank_cnt_1m—N0.0078PASSPASSFAIL——IV<0.02
bank_cnt_3m—N0.0260PASSPASSPASSPASS0.0482重要性排名超出Top-80
bank_cnt_6m—N0.0129PASSPASSFAIL——IV<0.02
bank_cnt_12m—N0.0208PASSPASSPASSFAIL0.0758与 loanlaworg 相关>0.7
c_bank_cnt_1m—N0.0079PASSPASSFAIL——IV<0.02
c_bank_cnt_3m—N0.0199PASSPASSFAIL——IV<0.02
c_bank_cnt_6m—N0.0130PASSPASSFAIL——IV<0.02
c_bank_cnt_12m—N0.0208PASSPASSPASSFAIL0.0791与 loanlaworg 相关>0.7
total_cnt_1m—N0.0208PASSPASSPASSFAIL0.0715与 nosvaing_bank_cnt_1m 相关>0.7
total_cnt_3m—N0.0279PASSPASSPASSFAIL0.0646与 nosvaing_bank_org_3m 相关>0.7
total_cnt_6m—N0.0108PASSPASSFAIL——IV<0.02
total_cnt_12m—N0.0196PASSPASSFAIL——IV<0.02
nosvaing_bank_cnt_1m—Y0.0508PASSPASSPASSPASS0.1895
nosvaing_bank_cnt_3m—N0.0287PASSPASSPASSFAIL0.1069与 nosvaing_bank_org_3m 相关>0.7
nosvaing_bank_cnt_6m—N0.0108PASSPASSFAIL——IV<0.02
nosvaing_bank_cnt_12m—N0.0150PASSPASSFAIL——IV<0.02
mini_loan_cnt_1m—N0.0003PASSPASSFAIL——IV<0.02
mini_loan_cnt_3m—N0.0022PASSPASSFAIL——IV<0.02
mini_loan_cnt_6m—N0.0105PASSPASSFAIL——IV<0.02
mini_loan_cnt_12m—N0.0220PASSPASSPASS—0.6091PSI>0.25
no_bank_997_cnt_1m—N0.0224PASSPASSPASSFAIL0.1614与 total_997_cnt_1m 相关>0.7
no_bank_997_cnt_3m—N0.0407PASSPASSPASSFAIL0.1242与 sum_month3_new_loan 相关>0.7
no_bank_997_cnt_6m—N0.0235PASSPASSPASSFAIL0.0086与 nosvaing_997_bank_cnt_6m 相关>0.7
no_bank_997_cnt_12m—N0.0109PASSPASSFAIL——IV<0.02
c_bank_997_cnt_1m—N0.0146PASSPASSFAIL——IV<0.02
c_bank_997_cnt_3m—N0.0186PASSPASSFAIL——IV<0.02
c_bank_997_cnt_6m—N0.0379PASSPASSPASSFAIL0.1196与 bank_997_cnt_6m 相关>0.7
c_bank_997_cnt_12m—N0.0237PASSPASSPASSFAIL0.1046与 bank_997_cnt_6m 相关>0.7
bank_997_cnt_1m—N0.0143PASSPASSFAIL——IV<0.02
bank_997_cnt_3m
近3月银行查询次数
—N0.0184PASSPASSFAIL——IV<0.02
bank_997_cnt_6m—Y0.0484PASSPASSPASSPASS0.1155
bank_997_cnt_12m—N0.0207PASSPASSPASSFAIL0.1015与 bank_997_cnt_6m 相关>0.7
total_997_cnt_1m—Y0.0321PASSPASSPASSPASS0.0760
total_997_cnt_3m—N0.0549PASSPASSPASSFAIL0.0718与 sum_month3_new_loan 相关>0.7
total_997_cnt_6m—N0.0377PASSPASSPASSFAIL0.0190与 bank_997_cnt_6m 相关>0.7
total_997_cnt_12m—N0.0304PASSPASSPASSFAIL0.0226与 bank_997_cnt_6m 相关>0.7
nosvaing_997_bank_cnt_1m—N0.0249PASSPASSPASSFAIL0.1870与 total_997_cnt_1m 相关>0.7
nosvaing_997_bank_cnt_3m—N0.0510PASSPASSPASSFAIL0.1890与 sum_month3_new_loan 相关>0.7
nosvaing_997_bank_cnt_6m—Y0.0347PASSPASSPASSPASS0.0234
nosvaing_997_bank_cnt_12m—N0.0206PASSPASSPASSFAIL0.0048与 nosvaing_997_bank_cnt_6m 相关>0.7
mini_997_loan_cnt_1m—N0.0013PASSFAIL———集中度>0.9
mini_997_loan_cnt_3m—N0.0007PASSPASSFAIL——IV<0.02
mini_997_loan_cnt_6m—N0.0049PASSPASSFAIL——IV<0.02
mini_997_loan_cnt_12m—N0.0092PASSPASSFAIL——IV<0.02
rh_m1_ln_qy_num—N0.0220PASSPASSPASSFAIL0.0703与 nosvaing_bank_cnt_1m 相关>0.7
rh_m1_crc_qy_num—N0.0000PASSFAIL———集中度>0.9
rh_guar_qy_num_m1—N0.0036PASSPASSFAIL——IV<0.02
c_bank_ln_uc_num—N0.0035PASSPASSFAIL——IV<0.02
bank_ln_uc_num—N0.0035PASSPASSFAIL——IV<0.02
no_bank_ln_uc_num—N0.0138PASSPASSFAIL——IV<0.02
nosvaing_bank_ln_uc_num—N0.0150PASSPASSFAIL——IV<0.02
mini_loan_ln_uc_num—N0.0000PASSFAIL———集中度>0.9
c_bank_ln_c_num—N0.0559PASSPASSPASSFAIL0.0431与 loannum 相关>0.7
bank_ln_c_num—N0.0550PASSPASSPASSFAIL0.0410与 loannum 相关>0.7
no_bank_ln_c_num—N0.0452PASSPASSPASSFAIL0.0416与 loannum 相关>0.7
nosvaing_bank_ln_c_num—N0.0315PASSPASSPASSFAIL0.0225与 loannum 相关>0.7
mini_loan_ln_c_num—N0.0272PASSPASSPASSPASS0.0617重要性排名超出Top-80
no_bank_998_rate_1m—N0.0250PASSPASSPASSFAIL0.1611与 nosvaing_bank_rate_1m 相关>0.7
no_bank_998_rate_3m—N0.0283PASSPASSPASSFAIL0.1803与 total_998_rate_3m 相关>0.7
no_bank_998_rate_6m—N0.0172PASSPASSFAIL——IV<0.02
no_bank_998_rate_12m—Y0.0404PASSPASSPASSPASS0.2307
c_bank_998_rate_1m—N0.0078PASSPASSFAIL——IV<0.02
c_bank_998_rate_3m—N0.0224PASSPASSPASSFAIL0.0287与 bank_rate_1_3 相关>0.7
c_bank_998_rate_6m—N0.0273PASSPASSPASSFAIL0.0486与 c_bank_rate_1_6 相关>0.7
c_bank_998_rate_12m—N0.0157PASSPASSFAIL——IV<0.02
bank_998_rate_1m—N0.0077PASSPASSFAIL——IV<0.02
bank_998_rate_3m—N0.0204PASSPASSPASSFAIL0.0255与 bank_rate_1_3 相关>0.7
bank_998_rate_6m—N0.0257PASSPASSPASSFAIL0.0455与 c_bank_rate_1_6 相关>0.7
bank_998_rate_12m—N0.0238PASSPASSPASSFAIL0.0286与 c_bank_rate_6_12 相关>0.7
total_998_rate_1m—N0.0252PASSPASSPASSFAIL0.0965与 nosvaing_bank_rate_1m 相关>0.7
total_998_rate_3m—Y0.0481PASSPASSPASSPASS0.0875
total_998_rate_6m—N0.0274PASSPASSPASSFAIL0.0553与 mini_loan_rate_6m 相关>0.7
total_998_rate_12m—N0.0224PASSPASSPASSFAIL0.1077与 loan_rate_1_12 相关>0.7
nosvaing_998_bank_rate_1m—N0.0333PASSPASSPASSFAIL0.1965与 nosvaing_bank_cnt_1m 相关>0.7
nosvaing_998_bank_rate_3m—N0.0411PASSPASSPASSFAIL0.1379与 nosvaing_bank_rate_1_3 相关>0.7
nosvaing_998_bank_rate_6m—Y0.0330PASSPASSPASSPASS0.0619
nosvaing_998_bank_rate_12m—N0.0263PASSPASSPASSPASS0.0643重要性排名超出Top-80
mini_998_loan_rate_1m—N0.0037PASSPASSFAIL——IV<0.02
mini_998_loan_rate_3m—N0.0094PASSPASSFAIL——IV<0.02
mini_998_loan_rate_6m—N0.0234PASSPASSPASS—0.3326PSI>0.25
mini_998_loan_rate_12m—N0.0261PASSPASSPASS—0.4285PSI>0.25
no_bank_rate_1m—N0.0226PASSPASSPASSFAIL0.0824与 nosvaing_bank_rate_1m 相关>0.7
no_bank_rate_3m—N0.0174PASSPASSFAIL——IV<0.02
no_bank_rate_6m—N0.0320PASSPASSPASSFAIL0.0614与 mini_loan_rate_6m 相关>0.7
no_bank_rate_12m—N0.0377PASSPASSPASSFAIL0.1365与 loan_rate_1_12 相关>0.7
c_bank_rate_1m—N0.0230PASSPASSPASSFAIL0.0784与 nosvaing_bank_rate_1m 相关>0.7
c_bank_rate_3m—N0.0169PASSPASSFAIL——IV<0.02
c_bank_rate_6m—N0.0210PASSPASSPASSFAIL0.0390与 mini_loan_rate_6m 相关>0.7
c_bank_rate_12m—N0.0168PASSPASSFAIL——IV<0.02
bank_rate_1m—N0.0226PASSPASSPASSFAIL0.0824与 nosvaing_bank_rate_1m 相关>0.7
bank_rate_3m—N0.0180PASSPASSFAIL——IV<0.02
bank_rate_6m—N0.0287PASSPASSPASSFAIL0.0593与 mini_loan_rate_6m 相关>0.7
bank_rate_12m—N0.0290PASSPASSPASSFAIL0.1339与 loan_rate_1_12 相关>0.7
nosvaing_bank_rate_1m—Y0.0434PASSPASSPASSPASS0.2355
nosvaing_bank_rate_3m—N0.0417PASSPASSPASSFAIL0.2003与 total_998_rate_3m 相关>0.7
nosvaing_bank_rate_6m—N0.0219PASSPASSPASSFAIL0.0715与 mini_loan_rate_6m 相关>0.7
nosvaing_bank_rate_12m—N0.0251PASSPASSPASSFAIL0.1564与 loan_rate_1_12 相关>0.7
mini_loan_rate_1m—N0.0382PASSPASSPASSFAIL0.1385与 nosvaing_bank_rate_1m 相关>0.7
mini_loan_rate_3m—N0.0340PASSPASSPASSFAIL0.1252与 total_998_rate_3m 相关>0.7
mini_loan_rate_6m—Y0.0374PASSPASSPASSPASS0.2120
mini_loan_rate_12m—N0.0349PASSPASSPASS—0.2933PSI>0.25
no_bank_rate_1_3—N0.0221PASSPASSPASSFAIL0.1193与 total_998_rate_3m 相关>0.7
no_bank_rate_1_6—N0.0315PASSPASSPASSFAIL0.2114与 no_bank_rate_3_6 相关>0.7
no_bank_rate_1_12—N0.0240PASSPASSPASS—0.2849PSI>0.25
no_bank_rate_3_6—Y0.0318PASSPASSPASSPASS0.1758
no_bank_rate_3_12—N0.0268PASSPASSPASSFAIL0.2349与 no_bank_998_rate_12m 相关>0.7
no_bank_rate_6_12—N0.0135PASSPASSFAIL——IV<0.02
c_bank_rate_1_3—N0.0201PASSPASSPASSFAIL0.0309与 bank_rate_1_3 相关>0.7
c_bank_rate_1_6—Y0.0281PASSPASSPASSPASS0.0431
c_bank_rate_1_12—N0.0163PASSPASSFAIL——IV<0.02
c_bank_rate_3_6—N0.0178PASSPASSFAIL——IV<0.02
c_bank_rate_3_12—N0.0112PASSPASSFAIL——IV<0.02
c_bank_rate_6_12—Y0.0394PASSPASSPASSPASS0.0194
bank_rate_1_3—N0.0254PASSPASSPASSPASS0.0353重要性排名超出Top-80
bank_rate_1_6—N0.0263PASSPASSPASSFAIL0.0589与 c_bank_rate_1_6 相关>0.7
bank_rate_1_12—N0.0177PASSPASSFAIL——IV<0.02
bank_rate_3_6—N0.0197PASSPASSFAIL——IV<0.02
bank_rate_3_12—N0.0144PASSPASSFAIL——IV<0.02
bank_rate_6_12—N0.0331PASSPASSPASSFAIL0.0170与 c_bank_rate_6_12 相关>0.7
nosvaing_bank_rate_1_3—Y0.0466PASSPASSPASSPASS0.2110
nosvaing_bank_rate_1_6—N0.0433PASSPASSPASS—0.2641PSI>0.25
nosvaing_bank_rate_1_12—N0.0401PASSPASSPASS—0.3052PSI>0.25
nosvaing_bank_rate_3_6—N0.0318PASSPASSPASSFAIL0.2487与 nosvaing_998_bank_rate_6m 相关>0.7
nosvaing_bank_rate_3_12—N0.0359PASSPASSPASS—0.3122PSI>0.25
nosvaing_bank_rate_6_12—N0.0242PASSPASSPASSFAIL0.1490与 nosvaing_998_bank_rate_12m 相关>0.7
mini_loan_rate_1_3—N0.0027PASSPASSFAIL——IV<0.02
mini_loan_rate_1_6—N0.0128PASSPASSFAIL——IV<0.02
mini_loan_rate_1_12—N0.0071PASSPASSFAIL——IV<0.02
mini_loan_rate_3_6—N0.0169PASSPASSFAIL——IV<0.02
mini_loan_rate_3_12—N0.0129PASSPASSFAIL——IV<0.02
mini_loan_rate_6_12—N0.0200PASSPASSFAIL——IV<0.02
loan_rate_1_3—N0.0299PASSPASSPASSFAIL0.1083与 total_998_rate_3m 相关>0.7
loan_rate_1_6—N0.0269PASSPASSPASSFAIL0.2047与 mini_loan_rate_6m 相关>0.7
loan_rate_1_12—Y0.0479PASSPASSPASSPASS0.2032
loan_rate_3_6—N0.0313PASSPASSPASSFAIL0.1544与 mini_loan_rate_6m 相关>0.7
loan_rate_3_12—N0.0404PASSPASSPASSFAIL0.1667与 loan_rate_1_12 相关>0.7
loan_rate_6_12—N0.0222PASSPASSPASSFAIL0.0419与 loan_rate_1_12 相关>0.7
card_rate_1_3—N0.0037PASSPASSFAIL——IV<0.02
card_rate_1_6—N0.0027PASSPASSFAIL——IV<0.02
card_rate_1_12—N0.0033PASSPASSFAIL——IV<0.02
card_rate_3_6—N0.0045PASSPASSFAIL——IV<0.02
card_rate_3_12—N0.0054PASSPASSFAIL——IV<0.02
card_rate_6_12—N0.0052PASSPASSFAIL——IV<0.02
loan_new_cnt_rate_3m—N0.0257PASSPASSPASSFAIL0.0734与 total_new_cnt_rate_3m 相关>0.7
loan_new_cnt_rate_6m—N0.0250PASSPASSPASSFAIL0.0730与 total_new_cnt_rate_3m 相关>0.7
loan_new_cnt_rate_12m—N0.0238PASSPASSPASSFAIL0.1673与 total_new_cnt_rate_3m 相关>0.7
card_new_cnt_rate_3m—N0.0144PASSPASSFAIL——IV<0.02
card_new_cnt_rate_6m—N0.0144PASSPASSFAIL——IV<0.02
card_new_cnt_rate_12m—N0.0267PASSPASSPASSFAIL0.0627与 c_bank_credit_used_limit 相关>0.7
total_new_cnt_rate_3m—Y0.0377PASSPASSPASSPASS0.0859
total_new_cnt_rate_6m—N0.0194PASSPASSFAIL——IV<0.02
total_new_cnt_rate_12m—N0.0372PASSPASSPASSFAIL0.1920与 total_new_cnt_rate_3m 相关>0.7
no_bank_cyc_used_limit—Y0.0278PASSPASSPASSPASS0.1613
no_bank_uncyc_used_limit—N0.0230PASSPASSPASS—0.2654PSI>0.25
no_bank_total_used_limit—Y0.0393PASSPASSPASSPASS0.0505
c_bank_cyc_used_limit—N0.0209PASSPASSPASSFAIL0.0075与 bank_cyc_used_limit 相关>0.7
c_bank_uncyc_used_limit—N0.0132PASSPASSFAIL——IV<0.02
c_bank_credit_used_limit—Y0.0291PASSPASSPASSPASS0.0686
c_bank_total_used_limit—N0.0380PASSPASSPASSFAIL0.0481与 bank_total_used_limit 相关>0.7
bank_cyc_used_limit—N0.0259PASSPASSPASSPASS0.0124重要性排名超出Top-80
bank_uncyc_used_limit—N0.0141PASSPASSFAIL——IV<0.02
bank_credit_used_limit—N0.0285PASSPASSPASSFAIL0.0679与 c_bank_credit_used_limit 相关>0.7
bank_total_used_limit—Y0.0426PASSPASSPASSPASS0.0440

06 回归结果

待补充:本实验未运行 LR 路线(或未产生回归结果),无回归明细。

07 VIF

待补充:本实验未运行 LR 路线,未计算 VIF/相关系数。

08 最终评分卡

待补充:本实验未生成评分卡(未运行 LR 路线)。

09 模型表现

模型区分度评估(按样本)

模型样本KSAUCGINI
LightGBM开发样本0.27200.66190.3238
LightGBM测试样本0.28210.66220.3244
LightGBM训练+测试样本0.27600.66200.3240
LightGBM跨时间验证样本0.23130.62380.2476

LightGBM — 4 样本评估表(score_cut 粒度)

开发样本(KS 0.2720,AUC 0.6619)

cutscntbadgoodbad_ratepopulation_pctcum_bad_ratecum_good_rate
[-inf, 231)4348407127793.63%63.40%66.22%39.01%
[231, 245)5394736687.76%7.86%73.91%48.31%
[245, 250)6035079684.08%8.79%82.16%61.83%
[250, inf)1368109727180.19%19.95%100.00%100.00%
2026-11-03T11:57:50.874576 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-11-03T11:57:50.980691 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

测试样本(KS 0.2821,AUC 0.6622)

cutscntbadgoodbad_ratepopulation_pctcum_bad_ratecum_good_rate
[-inf, 231)2860268417693.85%62.58%65.50%37.29%
[231, 245)3292943589.36%7.20%72.67%44.70%
[245, 250)4663868082.83%10.20%82.09%61.65%
[250, inf)91573418180.22%20.02%100.00%100.00%
2026-11-03T11:57:51.086512 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-11-03T11:57:51.190182 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

训练+测试样本(KS 0.2760,AUC 0.6620)

cutscntbadgoodbad_ratepopulation_pctcum_bad_ratecum_good_rate
[-inf, 231)7208675545393.72%63.07%65.93%38.32%
[231, 245)86876710188.36%7.60%73.41%46.87%
[245, 250)106989317683.54%9.35%82.13%61.76%
[250, inf)2283183145280.20%19.98%100.00%100.00%
2026-11-03T11:57:51.296676 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-11-03T11:57:51.400982 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

跨时间验证样本(KS 0.2313,AUC 0.6238)

cutscntbadgoodbad_ratepopulation_pctcum_bad_ratecum_good_rate
[-inf, 231)6786176191.00%63.66%66.70%43.57%
[231, 245)62501280.65%5.82%72.11%52.14%
[245, 250)101871486.14%9.48%81.51%62.14%
[250, inf)2241715376.34%21.03%100.00%100.00%
2026-11-03T11:57:51.507586 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-11-03T11:57:51.613624 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

跨时间验证样本结论

风险高度集中在低分尾端——最低 20% 样本集中了全量 66.7% 的坏客户,策略切分空间很大。

① 样本与分箱结构:共 1,065 样本 / 925 坏样本,整体坏率 86.85%,好坏比约 0.2:1;4 档等频分数分栏,每箱约 10%,个别箱占比偏薄/偏厚(5.8%~63.7%)。

② 排序性与单调性:逐箱 Bad Rate 91.00% → … → 76.34%(低分 → 高分),存在局部倒挂,建议复核分箱边界,最低分段坏率是最高分段的 1.2 倍。

③ 区分度与坏客户捕获:最低 20% 样本捕获全量 66.7% 的坏客户,最低 50% 捕获 66.7%;最高 20% 样本仅含 0.0% 的坏客户,其中风险最高一箱单独占 66.7%。

(口径:分数 = 按基准分/优比/PDO 刻度由模型概率转换,分数越高风险越低;结论由程序按分箱数据自动生成。)

注:score_cut 表按开发样本(train)分数等频 10 档取边界(低分 → 高分),各样本共用同一套边界、同栏可比(20261022 口径);分数由模型概率按基准分/优比/PDO 刻度转换,分数越高风险越低;每表附 KS/AUC(按概率计算,与分栏口径无关);训练+测试 = TRAIN 全体。旧实验无分数转换时为 PD 等频口径(按本样本分位切分)。

10 KS和GINI

LightGBM

样本KSGINIAUC
开发样本0.27200.32380.6619
测试样本0.28210.32440.6622
训练+测试样本0.27600.32400.6620
跨时间验证样本0.23130.24760.6238

开发样本

2026-11-03T11:57:51.694405 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-11-03T11:57:51.834790 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

测试样本

2026-11-03T11:57:51.902837 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-11-03T11:57:51.971492 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

训练+测试样本

2026-11-03T11:57:52.039381 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-11-03T11:57:52.107605 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

跨时间验证样本

2026-11-03T11:57:52.176007 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-11-03T11:57:52.244354 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

注:GINI = 2×AUC−1。KS 图 X 轴为分数(升序),红虚线为最大间隔位置;ROC 图例标注 AUC。

11 PSI

模型稳定性评估

模型训练集与测试集训练集与跨时间验证训练集与psi验证集
LightGBM0.00290.00860.0384

分数段 PSI 明细(开发样本 vs 跨时间验证样本)

分数段开发样本占比跨时间验证样本占比PSI 贡献累计 PSI
[-inf, 231)57.48%56.43%0.00020.0002
[231, 245)11.48%10.80%0.00040.0006
[245, 250)7.77%7.89%0.00000.0006
[250, inf)23.27%24.88%0.00110.0017

分数段 PSI 明细(开发样本 vs PSI时点样本)

分数段开发样本占比PSI时点占比PSI 贡献累计 PSI
[-inf, 231)57.48%66.86%0.01420.0142
[231, 245)11.48%9.22%0.00490.0191
[245, 250)7.77%3.75%0.02940.0485
[250, inf)23.27%20.17%0.00440.0529
2026-11-03T11:57:52.313363 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

注:分数段 PSI 明细按分数刻度计算(概率→分数转换,20261022 口径),分栏边界取自开发样本(train)分数等频分位,各对比样本共用同一套边界。

模型样本量分布评估

评分段开发样本样本量开发样本占比测试样本样本量测试样本占比训练+测试样本样本量训练+测试样本占比跨时间验证样本样本量跨时间验证样本占比PSI时点样本样本量PSI时点样本占比
(-inf, 231]394257.48%257956.43%652157.06%60156.43%46466.86%
(231, 245]78711.48%50811.12%129511.33%11510.80%649.22%
(245, 250]5337.77%3818.34%9148.00%847.89%263.75%
(250, inf]159623.27%110224.11%269823.61%26524.88%14020.17%
sum6858100.00%4570100.00%11428100.00%1065100.00%694100.00%

变量 PSI(开发样本 vs PSI时点样本,按入模变量分箱)

特征项PSI状态
ps_od_norm_int_amt0.2487关注
nosvaing_bank_rate_1m0.2355关注
repay_amount_m6_loan_os_prcp_pc0.2336关注
no_bank_998_rate_12m0.2307关注
setl_int_avg0.2302关注
sum_orig_prcp0.2245关注
apply_num_m120.2194关注
end_loan_to_cur_freq0.2156关注
loanearliestmonthsubcurmonth0.2133关注
mini_loan_rate_6m0.2120关注
nosvaing_bank_rate_1_30.2110关注
od_amt_m12_loan_os_prcp_m12_pc0.2101关注
loan_rate_1_120.2032关注
sum_apply_amt0.2029关注
useratemore90monthsubcurmonth0.1954关注
max_setl_rem_prcp_paym_m6_last_month_loan_os_prcp_pc0.1923关注
nosvaing_bank_cnt_1m0.1895关注
loan_os_prcp_orig_prcp_pc0.1869关注
setl_prcp_avg0.1761关注
no_bank_rate_3_60.1758关注
sum_open_loan_mob24_neverdur0.1749关注
setl_rem_prcp_paym_m3_loan_os_prcp_pc0.1715关注
repay_amount_last_month_loan_os_prcp_pc0.1697关注
apply_amt_avg_m120.1669关注
no_bank_cyc_used_limit0.1613关注
repay_amount_m12_all_shd_amt_pc0.1581关注
all_repay_amount_m3_shd_amt_pc0.1558关注
cardaveragemob0.1555关注
loan_rate_3m_12m0.1493关注
cur_od_loan_amt0.1478关注
allcardsoverdraftrate6avg0.1372关注
account_age_loan_day0.1271关注
repay_amount_all_shd_amt_pc0.1174关注
bank_997_cnt_6m0.1155关注
max_setl_rem_prcp_paym_m3_last_month_loan_os_prcp_pc0.1148关注
loanrecentlymonthsubcurmonth0.1078关注
mostcarduserate0.1039关注
nosvaing_bank_org_3m0.1033关注
acc_ps_perd_no0.1017关注
normal_repay_num0.1011关注
orig_prcp_avg_m60.0921稳定
sum_card_mob12_neverdur0.0898稳定
total_998_rate_3m0.0875稳定
total_new_cnt_rate_3m0.0859稳定
sum_month3_new_loan0.0811稳定
total_997_cnt_1m0.0760稳定
ovd_days_d15_num_m30.0731稳定
c_bank_credit_used_limit0.0686稳定
loan_reject_rate_3m0.0683稳定
odgt0_m12_mob_pc0.0678稳定
nosvaing_998_bank_rate_6m0.0619稳定
odgt0_m12_num0.0597稳定
useratemore90_m6_num0.0585稳定
loan_reject_rate_6m0.0550稳定
account_age0.0525稳定
loannum0.0509稳定
no_bank_total_used_limit0.0505稳定
query_success_rate_6m0.0489稳定
f_ps_od_days_max0.0471稳定
last6average0.0458稳定
bank_total_used_limit0.0440稳定
c_bank_rate_1_60.0431稳定
setl_prcp_mon_ratio0.0399稳定
apply_num_m60.0347稳定
setl_rem_prcp_paym_m12_loan_os_prcp_pc0.0344稳定
setl_int_mon0.0322稳定
setl_rem_prcp_paym_m3_num0.0322稳定
tothismonthodcontiaddnumm120.0301稳定
ovd_days_d30_num_m60.0296稳定
max_od_days_m120.0278稳定
nosvaing_997_bank_cnt_6m0.0234稳定
ncarloanbalancerate6month0.0213稳定
c_bank_rate_6_120.0194稳定
od_amt_m3_loan_os_prcp_m3_pc0.0184稳定
m_c_over_due_cnt0.0162稳定
rh_ln_qy_fst_inte0.0118稳定
t_ps_od_days_max0.0110稳定
rh_cr_bal_pro0.0065稳定
cyc_accountcount0.0049稳定
odduecontireducenumm120.0040稳定
odgt0_m12_num — PSI 0.0597
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 1.5]43.80%55.48%11.67%0.0276
(1.5, 2.5]21.45%16.71%-4.73%0.0118
(2.5, 3.5]11.90%9.51%-2.39%0.0054
(3.5, 6.5]16.73%12.10%-4.62%0.0149
(6.5, inf]6.12%6.20%0.07%0.0000
odgt0_m12_mob_pc — PSI 0.0678
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 23.61]45.23%51.30%6.07%0.0076
(23.61, 39.75]24.09%17.72%-6.37%0.0195
(39.75, 58.79]17.42%13.69%-3.74%0.0090
(58.79, 73.87]5.75%4.47%-1.28%0.0032
(73.87, inf]7.51%12.82%5.31%0.0284
m_c_over_due_cnt — PSI 0.0162
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 1.5]68.96%74.50%5.54%0.0043
(1.5, 2.5]17.19%13.54%-3.65%0.0087
(2.5, 3.5]6.58%5.33%-1.24%0.0026
(3.5, inf]7.28%6.63%-0.65%0.0006
acc_ps_perd_no — PSI 0.1017
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 2.5]44.39%59.37%14.98%0.0436
(2.5, 5.5]26.42%20.61%-5.82%0.0145
(5.5, 6.5]5.12%4.61%-0.51%0.0005
(6.5, 16.5]16.84%11.82%-5.03%0.0178
(16.5, inf]7.23%3.60%-3.63%0.0253
tothismonthodcontiaddnumm12 — PSI 0.0301
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]43.63%35.16%-8.47%0.0183
(0.5, inf]56.37%64.84%8.47%0.0119
loan_os_prcp_orig_prcp_pc — PSI 0.1869
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 13.04]8.02%5.91%-2.11%0.0065
(13.04, 53.4]48.72%35.01%-13.70%0.0453
(53.4, 67.97]15.46%10.81%-4.65%0.0166
(67.97, 76.46]8.15%11.53%3.38%0.0117
(76.46, inf]19.66%36.74%17.09%0.1069
all_repay_amount_m3_shd_amt_pc — PSI 0.1558
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 66.67]45.73%56.63%10.90%0.0233
(66.67, 66.73]5.50%5.04%-0.45%0.0004
(66.73, 69.51]9.00%10.95%1.95%0.0038
(69.51, 99.99]10.06%13.40%3.34%0.0096
(99.99, inf]29.72%13.98%-15.74%0.1187
ovd_days_d30_num_m6 — PSI 0.0296
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]76.13%83.00%6.87%0.0059
(0.5, 1.5]16.45%11.38%-5.06%0.0186
(1.5, inf]7.42%5.62%-1.80%0.0050
od_amt_m3_loan_os_prcp_m3_pc — PSI 0.0184
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 10.09]71.49%75.94%4.44%0.0027
(10.09, 14.35]10.44%9.51%-0.93%0.0009
(14.35, 24.15]7.92%4.90%-3.02%0.0145
(24.15, 25.53]5.16%4.75%-0.41%0.0003
(25.53, inf]4.99%4.90%-0.09%0.0000
repay_amount_last_month_loan_os_prcp_pc — PSI 0.1697
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.79]66.46%75.65%9.19%0.0119
(0.79, 8.66]5.02%10.66%5.65%0.0426
(8.66, 20.06]11.94%4.61%-7.33%0.0698
(20.06, 49.49]8.78%4.61%-4.17%0.0268
(49.49, inf]7.80%4.47%-3.33%0.0186
max_od_days_m12 — PSI 0.0278
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 4.5]6.62%7.20%0.58%0.0005
(4.5, 7.5]7.80%8.21%0.41%0.0002
(7.5, 9.5]5.95%5.48%-0.47%0.0004
(9.5, 29.5]66.77%71.18%4.41%0.0028
(29.5, inf]12.86%7.93%-4.94%0.0239
ps_od_norm_int_amt — PSI 0.2487
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 10.25]7.01%4.75%-2.26%0.0088
(10.25, 44.34]21.71%15.42%-6.29%0.0215
(44.34, 107.69]22.65%13.40%-9.24%0.0485
(107.69, 757.16]41.25%43.95%2.70%0.0017
(757.16, inf]7.38%22.48%15.10%0.1682
repay_amount_all_shd_amt_pc — PSI 0.1174
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 33.16]66.13%75.07%8.94%0.0113
(33.16, 84.39]5.34%8.79%3.45%0.0172
(84.39, 90.67]11.23%4.32%-6.91%0.0659
(90.67, 96.6]9.86%6.05%-3.81%0.0186
(96.6, inf]7.45%5.76%-1.69%0.0043
end_loan_to_cur_freq — PSI 0.2156
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, -362.5]6.12%6.48%0.36%0.0002
(-362.5, -248.5]25.75%13.54%-12.21%0.0784
(-248.5, -109.5]27.37%17.58%-9.79%0.0433
(-109.5, -26.5]8.19%11.24%3.04%0.0096
(-26.5, inf]32.56%51.15%18.59%0.0840
cur_od_loan_amt — PSI 0.1478
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 124.01]8.34%6.77%-1.57%0.0033
(124.01, 558.05]28.55%21.18%-7.37%0.0220
(558.05, 725.59]6.49%5.19%-1.30%0.0029
(725.59, 3545.66]43.92%39.34%-4.58%0.0050
(3545.66, inf]12.70%27.52%14.82%0.1146
odduecontireducenumm12 — PSI 0.0040
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]74.39%77.09%2.69%0.0010
(0.5, inf]25.61%22.91%-2.69%0.0030
max_setl_rem_prcp_paym_m3_last_month_loan_os_prcp_pc — PSI 0.1148
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 1.88]69.92%77.67%7.75%0.0081
(1.88, 11.68]7.93%11.53%3.60%0.0134
(11.68, 20.03]9.70%3.46%-6.24%0.0643
(20.03, 33.46]5.22%3.60%-1.62%0.0060
(33.46, inf]7.23%3.75%-3.49%0.0229
apply_amt_avg_m12 — PSI 0.1669
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 4525]43.80%29.83%-13.98%0.0537
(4525, 10810]17.02%18.44%1.43%0.0011
(10810, 32366.67]21.60%22.19%0.60%0.0002
(32366.67, 48025]7.32%6.05%-1.27%0.0024
(48025, inf]10.27%23.49%13.22%0.1094
orig_prcp_avg_m6 — PSI 0.0921
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 636.67]67.16%61.82%-5.35%0.0044
(636.67, 2158.33]6.85%4.75%-2.10%0.0077
(2158.33, 5016.67]7.68%5.33%-2.35%0.0086
(5016.67, 15012.5]9.20%9.22%0.02%0.0000
(15012.5, inf]9.10%18.88%9.78%0.0713
account_age_loan_day — PSI 0.1271
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.52]23.99%38.76%14.77%0.0709
(0.52, 0.96]32.06%29.39%-2.67%0.0023
(0.96, 1.19]9.42%9.37%-0.05%0.0000
(1.19, 4.62]21.19%12.82%-8.36%0.0420
(4.62, inf]13.34%9.65%-3.69%0.0119
repay_amount_m6_loan_os_prcp_pc — PSI 0.2336
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 42.01]31.99%54.03%22.04%0.1155
(42.01, 70.26]33.51%23.92%-9.59%0.0323
(70.26, 82.59]7.63%6.77%-0.85%0.0010
(82.59, 99.67]5.09%4.90%-0.19%0.0001
(99.67, inf]21.78%10.37%-11.41%0.0846
sum_orig_prcp — PSI 0.2245
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 10450]36.53%29.54%-6.99%0.0148
(10450, 12554]5.18%2.02%-3.16%0.0298
(12554, 19150]12.03%6.92%-5.11%0.0283
(19150, 110100]40.96%43.80%2.84%0.0019
(110100, inf]5.31%17.72%12.42%0.1497
loannum — PSI 0.0509
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 9.5]6.74%4.07%-2.67%0.0134
(9.5, 27.5]31.29%24.43%-6.86%0.0170
(27.5, 39.5]16.08%17.80%1.72%0.0017
(39.5, 61.5]18.78%19.00%0.23%0.0000
(61.5, inf]27.11%34.69%7.58%0.0187
sum_card_mob12_neverdur — PSI 0.0898
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]32.50%20.97%-11.54%0.0506
(0.5, 1.5]12.92%11.46%-1.46%0.0017
(1.5, 6.5]28.77%32.13%3.36%0.0037
(6.5, 8.5]5.19%5.73%0.54%0.0005
(8.5, inf]20.62%29.71%9.09%0.0332
od_amt_m12_loan_os_prcp_m12_pc — PSI 0.2101
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.91]18.10%34.44%16.34%0.1052
(0.91, 1.09]5.61%4.75%-0.86%0.0014
(1.09, 2.17]51.62%34.58%-17.04%0.0682
(2.17, 3.17]11.40%7.49%-3.91%0.0164
(3.17, inf]13.27%18.73%5.46%0.0188
setl_int_mon — PSI 0.0322
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 3.64]73.04%73.78%0.74%0.0001
(3.64, 17.71]5.02%3.46%-1.56%0.0058
(17.71, 42.33]5.40%3.75%-1.65%0.0060
(42.33, 90.6]5.50%3.75%-1.75%0.0067
(90.6, inf]11.05%15.27%4.22%0.0137
setl_rem_prcp_paym_m3_loan_os_prcp_pc — PSI 0.1715
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, -4999]5.06%15.13%10.07%0.1103
(-4999, 3.21]65.70%63.98%-1.73%0.0005
(3.21, 19.05]12.50%12.68%0.18%0.0000
(19.05, 25.49]5.32%2.59%-2.73%0.0196
(25.49, inf]11.42%5.62%-5.80%0.0411
sum_month3_new_loan — PSI 0.0811
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]21.94%12.67%-9.27%0.0509
(0.5, 1.5]20.57%18.55%-2.02%0.0021
(1.5, 2.5]16.86%22.47%5.62%0.0162
(2.5, 3.5]11.71%15.54%3.82%0.0108
(3.5, inf]28.92%30.77%1.85%0.0011
sum_apply_amt — PSI 0.2029
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 2050]7.48%5.48%-2.00%0.0063
(2050, 9625]25.90%15.85%-10.05%0.0493
(9625, 11585]5.69%3.89%-1.80%0.0068
(11585, 130400]55.66%57.64%1.98%0.0007
(130400, inf]5.28%17.15%11.87%0.1398
loan_reject_rate_3m — PSI 0.0683
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, -0.15]12.82%19.61%6.79%0.0288
(-0.15, 0.02]23.62%23.23%-0.39%0.0001
(0.02, 0.28]7.20%6.18%-1.01%0.0015
(0.28, 0.58]20.84%25.04%4.20%0.0077
(0.58, inf]35.52%25.94%-9.58%0.0301
setl_prcp_mon_ratio — PSI 0.0399
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0]72.76%79.83%7.07%0.0065
(0, 0.01]7.42%6.77%-0.65%0.0006
(0.01, 0.02]5.44%4.03%-1.40%0.0042
(0.02, 0.04]6.88%5.62%-1.26%0.0026
(0.04, inf]7.49%3.75%-3.75%0.0260
useratemore90_m6_num — PSI 0.0585
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]64.25%55.33%-8.91%0.0133
(0.5, 1.5]8.31%12.39%4.08%0.0163
(1.5, 2.5]12.54%14.55%2.01%0.0030
(2.5, 3.5]6.43%10.81%4.38%0.0227
(3.5, inf]8.47%6.92%-1.56%0.0032
account_age — PSI 0.0525
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 6.5]27.66%27.95%0.29%0.0000
(6.5, 11.5]36.66%41.35%4.70%0.0057
(11.5, 14.5]10.45%9.80%-0.66%0.0004
(14.5, 33.5]19.68%12.39%-7.29%0.0338
(33.5, inf]5.54%8.50%2.96%0.0127
apply_num_m6 — PSI 0.0347
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]65.33%56.77%-8.55%0.0120
(0.5, 1.5]20.98%26.80%5.82%0.0142
(1.5, 2.5]6.77%9.37%2.60%0.0085
(2.5, inf]6.93%7.06%0.13%0.0000
repay_amount_m12_all_shd_amt_pc — PSI 0.1581
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 85.33]17.25%30.69%13.44%0.0774
(85.33, 86.76]5.45%6.92%1.46%0.0035
(86.76, 89.01]6.98%10.81%3.82%0.0167
(89.01, 99.76]65.30%48.70%-16.59%0.0486
(99.76, inf]5.02%2.88%-2.13%0.0118
t_ps_od_days_max — PSI 0.0110
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]73.43%77.38%3.94%0.0021
(0.5, 3.5]11.24%8.79%-2.45%0.0060
(3.5, 11.5]6.30%6.34%0.04%0.0000
(11.5, inf]9.03%7.49%-1.53%0.0029
nosvaing_bank_cnt_1m — PSI 0.1895
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]65.57%44.67%-20.90%0.0802
(0.5, 1.5]21.73%30.84%9.11%0.0319
(1.5, inf]12.71%24.50%11.79%0.0774
last6average — PSI 0.0458
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 128.5]22.80%23.20%0.40%0.0001
(128.5, 977.5]19.68%13.26%-6.43%0.0254
(977.5, 2255.5]17.74%15.42%-2.33%0.0033
(2255.5, 3242.5]9.91%10.95%1.04%0.0010
(3242.5, inf]29.87%37.18%7.31%0.0160
useratemore90monthsubcurmonth — PSI 0.1954
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 1.5]13.02%17.72%4.70%0.0145
(1.5, 3.5]11.13%15.42%4.29%0.0140
(3.5, 5.5]11.61%11.53%-0.08%0.0000
(5.5, 8.5]14.23%25.36%11.13%0.0643
(8.5, inf]50.01%29.97%-20.04%0.1026
setl_int_avg — PSI 0.2302
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 13.19]14.98%13.69%-1.29%0.0012
(13.19, 16.74]4.99%2.45%-2.54%0.0180
(16.74, 22.2]6.88%3.46%-3.42%0.0236
(22.2, 110.95]47.93%33.43%-14.50%0.0522
(110.95, inf]25.23%46.97%21.75%0.1352
bank_997_cnt_6m — PSI 0.1155
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 1.5]49.98%62.97%12.98%0.0300
(1.5, 3.5]24.16%23.20%-0.96%0.0004
(3.5, 5.5]11.97%7.93%-4.05%0.0167
(5.5, 9.5]8.79%3.89%-4.90%0.0400
(9.5, inf]5.09%2.02%-3.07%0.0285
total_998_rate_3m — PSI 0.0875
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, -4999]8.59%6.20%-2.40%0.0078
(-4999, 0.1]14.45%18.59%4.14%0.0104
(0.1, 0.52]30.54%37.18%6.64%0.0131
(0.52, 0.73]18.44%20.89%2.45%0.0031
(0.73, inf]27.98%17.15%-10.83%0.0531
ovd_days_d15_num_m3 — PSI 0.0731
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]33.33%33.72%0.38%0.0000
(0.5, 1.5]40.38%50.00%9.62%0.0206
(1.5, 2.5]12.80%9.08%-3.72%0.0128
(2.5, 4.5]8.31%4.61%-3.70%0.0218
(4.5, inf]5.18%2.59%-2.58%0.0178
loan_rate_1_12 — PSI 0.2032
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.02]38.39%27.23%-11.16%0.0383
(0.02, 0.05]9.53%4.32%-5.20%0.0411
(0.05, 0.07]6.73%4.61%-2.12%0.0080
(0.07, 0.27]36.25%42.36%6.11%0.0095
(0.27, inf]9.10%21.47%12.37%0.1062
nosvaing_bank_rate_1_3 — PSI 0.2110
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, -4999]31.35%18.16%-13.19%0.0720
(-4999, 0.28]39.45%31.70%-7.75%0.0169
(0.28, 0.39]5.18%6.48%1.30%0.0029
(0.39, 0.74]12.22%20.61%8.39%0.0438
(0.74, inf]11.80%23.05%11.25%0.0753
allcardsoverdraftrate6avg — PSI 0.1372
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.07]48.71%47.51%-1.20%0.0003
(0.07, 0.88]29.37%42.08%12.71%0.0457
(0.88, 0.95]6.63%3.92%-2.71%0.0142
(0.95, 0.99]5.89%2.87%-3.02%0.0217
(0.99, inf]9.40%3.62%-5.78%0.0552
f_ps_od_days_max — PSI 0.0471
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]84.18%79.39%-4.78%0.0028
(0.5, 2.5]9.36%7.93%-1.44%0.0024
(2.5, inf]6.46%12.68%6.22%0.0420
setl_rem_prcp_paym_m12_loan_os_prcp_pc — PSI 0.0344
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, -9998.5]64.58%69.74%5.16%0.0040
(-9998.5, -4999]8.15%9.22%1.07%0.0013
(-4999, 17.22]11.74%11.38%-0.35%0.0001
(17.22, 79.79]9.27%6.20%-3.08%0.0124
(79.79, inf]6.26%3.46%-2.80%0.0166
cardaveragemob — PSI 0.1555
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 20.5]33.18%22.32%-10.86%0.0431
(20.5, 75.5]28.40%20.97%-7.43%0.0225
(75.5, 84.5]7.47%7.99%0.53%0.0004
(84.5, 104.5]13.59%17.80%4.21%0.0114
(104.5, inf]17.37%30.92%13.55%0.0782
sum_open_loan_mob24_neverdur — PSI 0.1749
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]11.00%5.13%-5.87%0.0448
(0.5, 3.5]48.31%34.99%-13.32%0.0430
(3.5, 4.5]11.46%14.03%2.56%0.0052
(4.5, 8.5]22.88%33.18%10.30%0.0383
(8.5, inf]6.35%12.67%6.32%0.0436
ncarloanbalancerate6month — PSI 0.0213
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.19]23.78%22.62%-1.15%0.0006
(0.19, 0.63]5.34%7.39%2.05%0.0066
(0.63, 2.04]53.52%56.41%2.89%0.0015
(2.04, 2.81]5.73%5.43%-0.30%0.0002
(2.81, inf]11.63%8.14%-3.49%0.0124
nosvaing_bank_rate_1m — PSI 0.2355
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, -4999]37.06%26.22%-10.84%0.0375
(-4999, 0.04]28.50%18.44%-10.06%0.0438
(0.04, 0.35]10.60%9.94%-0.65%0.0004
(0.35, 0.61]11.62%17.44%5.81%0.0236
(0.61, inf]12.22%27.95%15.74%0.1303
setl_prcp_avg — PSI 0.1761
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 88.54]10.50%13.11%2.61%0.0058
(88.54, 609.66]50.67%33.14%-17.53%0.0744
(609.66, 719.26]5.22%4.18%-1.04%0.0023
(719.26, 936.76]7.22%5.62%-1.60%0.0040
(936.76, inf]26.39%43.95%17.56%0.0895
bank_total_used_limit — PSI 0.0440
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.16]17.23%24.21%6.98%0.0237
(0.16, 0.21]4.68%5.33%0.65%0.0009
(0.21, 0.31]11.68%12.39%0.71%0.0004
(0.31, 0.75]46.09%43.23%-2.86%0.0018
(0.75, inf]20.32%14.84%-5.48%0.0172
rh_cr_bal_pro — PSI 0.0065
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, -4999]6.09%4.52%-1.56%0.0046
(-4999, 0.45]17.46%18.85%1.39%0.0011
(0.45, 0.82]46.77%45.70%-1.07%0.0002
(0.82, 0.97]22.88%23.83%0.95%0.0004
(0.97, inf]6.80%7.09%0.29%0.0001
loan_reject_rate_6m — PSI 0.0550
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.25]46.65%56.26%9.61%0.0180
(0.25, 0.51]24.34%23.68%-0.66%0.0002
(0.51, 0.62]7.37%6.18%-1.18%0.0021
(0.62, 0.69]6.88%4.22%-2.66%0.0130
(0.69, inf]14.76%9.65%-5.11%0.0217
no_bank_998_rate_12m — PSI 0.2307
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.22]16.83%34.29%17.46%0.1242
(0.22, 0.33]9.95%12.68%2.73%0.0066
(0.33, 0.68]58.61%46.69%-11.92%0.0271
(0.68, 0.74]5.28%1.73%-3.55%0.0396
(0.74, inf]9.33%4.61%-4.72%0.0332
max_setl_rem_prcp_paym_m6_last_month_loan_os_prcp_pc — PSI 0.1923
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 3.39]49.46%61.24%11.78%0.0252
(3.39, 9.1]7.49%14.84%7.35%0.0502
(9.1, 11.68]7.57%3.60%-3.97%0.0294
(11.68, 90.9]26.36%15.99%-10.37%0.0518
(90.9, inf]9.11%4.32%-4.79%0.0357
c_bank_rate_6_12 — PSI 0.0194
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.24]20.69%25.07%4.38%0.0084
(0.24, 0.39]12.87%14.70%1.82%0.0024
(0.39, 0.44]5.90%4.90%-1.00%0.0019
(0.44, 0.68]31.73%27.38%-4.35%0.0064
(0.68, inf]28.81%27.95%-0.85%0.0003
no_bank_total_used_limit — PSI 0.0505
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.27]53.24%63.11%9.87%0.0168
(0.27, 0.3]5.81%6.20%0.39%0.0002
(0.3, 0.33]5.73%4.75%-0.98%0.0018
(0.33, 0.58]25.84%20.17%-5.67%0.0140
(0.58, inf]9.37%5.76%-3.61%0.0176
mostcarduserate — PSI 0.1039
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.15]49.68%49.62%-0.06%0.0000
(0.15, 0.82]10.74%19.76%9.02%0.0550
(0.82, 0.99]12.18%13.42%1.25%0.0012
(0.99, 1.07]22.20%13.88%-8.32%0.0391
(1.07, inf]5.21%3.32%-1.89%0.0085
cyc_accountcount — PSI 0.0049
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 3.5]5.06%3.90%-1.16%0.0030
(3.5, 10.5]48.26%48.70%0.44%0.0000
(10.5, 11.5]6.39%7.51%1.12%0.0018
(11.5, 16.5]23.43%23.12%-0.31%0.0000
(16.5, inf]16.86%16.76%-0.09%0.0000
total_new_cnt_rate_3m — PSI 0.0859
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0]20.98%11.96%-9.02%0.0507
(0, 0.02]8.70%12.54%3.84%0.0140
(0.02, 0.11]48.85%56.34%7.49%0.0107
(0.11, 0.17]10.81%11.53%0.72%0.0005
(0.17, inf]10.66%7.64%-3.02%0.0101
mini_loan_rate_6m — PSI 0.2120
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.2]37.97%59.94%21.98%0.1004
(0.2, 0.37]34.27%23.34%-10.92%0.0419
(0.37, 0.48]11.30%5.33%-5.97%0.0448
(0.48, 0.57]8.30%6.92%-1.39%0.0025
(0.57, inf]8.17%4.47%-3.70%0.0223
apply_num_m12 — PSI 0.2194
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]22.59%7.64%-14.95%0.1621
(0.5, 1.5]45.36%51.01%5.65%0.0066
(1.5, 2.5]15.02%24.50%9.48%0.0464
(2.5, 3.5]6.39%7.64%1.25%0.0022
(3.5, inf]10.64%9.22%-1.42%0.0020
loanearliestmonthsubcurmonth — PSI 0.2133
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 47.5]10.92%3.17%-7.75%0.0960
(47.5, 53.5]5.58%1.96%-3.62%0.0378
(53.5, 85.5]31.14%25.49%-5.65%0.0113
(85.5, 93.5]10.36%9.05%-1.31%0.0018
(93.5, inf]42.00%60.33%18.33%0.0664
loan_rate_3m_12m — PSI 0.1493
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.07]11.13%6.94%-4.19%0.0198
(0.07, 0.29]33.33%21.27%-12.07%0.0542
(0.29, 0.34]10.07%11.16%1.09%0.0011
(0.34, 0.62]31.89%35.60%3.71%0.0041
(0.62, inf]13.58%25.04%11.46%0.0701
normal_repay_num — PSI 0.1011
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 30.5]73.02%85.30%12.28%0.0191
(30.5, 38.5]5.56%4.03%-1.52%0.0049
(38.5, 66.5]10.08%5.62%-4.46%0.0260
(66.5, 103.5]5.05%2.31%-2.74%0.0215
(103.5, inf]6.30%2.74%-3.56%0.0297
loanrecentlymonthsubcurmonth — PSI 0.1078
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 1.5]42.29%56.56%14.27%0.0415
(1.5, 3.5]36.16%31.07%-5.09%0.0077
(3.5, 4.5]8.27%4.52%-3.75%0.0226
(4.5, 6.5]8.13%3.77%-4.36%0.0335
(6.5, inf]5.14%4.07%-1.07%0.0025
setl_rem_prcp_paym_m3_num — PSI 0.0322
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]69.41%77.09%7.68%0.0081
(0.5, 1.5]19.15%15.27%-3.87%0.0087
(1.5, inf]11.45%7.64%-3.81%0.0154
nosvaing_997_bank_cnt_6m — PSI 0.0234
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]18.10%13.98%-4.13%0.0107
(0.5, 3.5]53.04%52.59%-0.45%0.0000
(3.5, 4.5]8.59%8.93%0.34%0.0001
(4.5, 6.5]9.40%9.65%0.25%0.0001
(6.5, inf]10.86%14.84%3.99%0.0125
query_success_rate_6m — PSI 0.0489
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.19]10.10%8.14%-1.96%0.0042
(0.19, 0.41]23.26%15.84%-7.43%0.0286
(0.41, 0.69]34.05%36.35%2.30%0.0015
(0.69, 0.77]6.65%7.39%0.74%0.0008
(0.77, inf]25.94%32.28%6.34%0.0139
rh_ln_qy_fst_inte — PSI 0.0118
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 675.5]42.19%42.23%0.05%0.0000
(675.5, 688.5]6.48%5.88%-0.59%0.0006
(688.5, 706.5]14.68%12.52%-2.16%0.0034
(706.5, 724.5]23.56%27.60%4.04%0.0064
(724.5, inf]13.09%11.76%-1.33%0.0014
nosvaing_998_bank_rate_6m — PSI 0.0619
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, -3.75]14.66%12.82%-1.84%0.0025
(-3.75, -1.13]6.36%5.33%-1.03%0.0018
(-1.13, 0.26]35.58%27.81%-7.77%0.0192
(0.26, 0.76]27.48%29.54%2.06%0.0015
(0.76, inf]15.92%24.50%8.58%0.0370
total_997_cnt_1m — PSI 0.0760
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]55.99%42.65%-13.34%0.0363
(0.5, 1.5]24.95%30.12%5.16%0.0097
(1.5, 2.5]10.05%15.13%5.08%0.0208
(2.5, inf]9.01%12.10%3.10%0.0092
no_bank_rate_3_6 — PSI 0.1758
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.16]17.25%12.25%-5.00%0.0171
(0.16, 0.3]10.75%3.75%-7.00%0.0738
(0.3, 0.36]7.26%5.04%-2.22%0.0081
(0.36, 0.53]21.68%17.29%-4.39%0.0099
(0.53, inf]43.06%61.67%18.61%0.0669
nosvaing_bank_org_3m — PSI 0.1033
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]31.35%18.16%-13.19%0.0720
(0.5, 1.5]27.05%28.53%1.48%0.0008
(1.5, 2.5]17.34%22.77%5.43%0.0148
(2.5, 4.5]16.77%22.05%5.27%0.0144
(4.5, inf]7.49%8.50%1.01%0.0013
c_bank_credit_used_limit — PSI 0.0686
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.09]48.84%48.13%-0.71%0.0001
(0.09, 0.73]32.55%40.78%8.22%0.0185
(0.73, 0.83]5.23%1.87%-3.36%0.0345
(0.83, 0.99]7.28%4.90%-2.38%0.0094
(0.99, inf]6.10%4.32%-1.78%0.0061
c_bank_rate_1_6 — PSI 0.0431
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.12]66.05%62.97%-3.09%0.0015
(0.12, 0.23]9.16%6.92%-2.24%0.0063
(0.23, 0.33]5.78%4.61%-1.17%0.0026
(0.33, 0.59]12.60%13.83%1.23%0.0012
(0.59, inf]6.41%11.67%5.26%0.0316
no_bank_cyc_used_limit — PSI 0.1613
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.24]62.39%80.26%17.87%0.0450
(0.24, 0.27]5.05%2.88%-2.16%0.0121
(0.27, 0.33]8.99%4.32%-4.67%0.0342
(0.33, 0.4]7.49%4.18%-3.31%0.0193
(0.4, inf]16.09%8.36%-7.73%0.0506

注:PSI ≤ 0.1 视为稳定,0.1~0.25 需关注,> 0.25 不稳定。

12 单变量分析

注:统计所有入模变量的单变量分析(开发样本口径)。预留验证/总体/OOT 口径的分组明细待扩展。

odgt0_m12_num
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 1.5]300443.80%169283594.37%0.06
2(1.5, 2.5]147165.25%126134591.43%0.09
3(2.5, 3.5]81677.15%11070686.52%0.16
4(3.5, 6.5]114793.88%19994882.65%0.21
5(6.5, inf]420100.00%10631474.76%0.34
合计6858710614889.65%0.12
2026-11-03T11:57:52.424966 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调下降(94.37% → 74.76%),WOE 单调上升(-0.6613 → 1.0727),符合风控分箱预期。IV=0.3489,该变量区分度尚可。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 1 箱((-inf, 1.5])为极高风险区间:坏率高达 94.37%,策略上应重点管控

odgt0_m12_mob_pc
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 23.61]310245.23%188291493.94%0.06
2(23.61, 39.75]165269.32%165148790.01%0.11
3(39.75, 58.79]119586.75%20499182.93%0.21
4(58.79, 73.87]39492.49%5134387.06%0.15
5(73.87, inf]515100.00%10241380.19%0.25
合计6858710614889.65%0.12
2026-11-03T11:57:52.557695 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(93.94% → 93.94% → 80.19%),如需严格单调可考虑调整切点或合并分箱。IV=0.2573,该变量区分度尚可。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 82.93% 回升至 87.06%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 23.61])为极高风险区间:坏率高达 93.94%,策略上应重点管控

m_c_over_due_cnt
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 1.5]472968.96%343438692.75%0.08
2(1.5, 2.5]117986.15%172100785.41%0.17
3(2.5, 3.5]45192.72%8436781.37%0.23
4(3.5, inf]499100.00%11138877.76%0.29
合计6858710614889.65%0.12
2026-11-03T11:57:52.684001 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调下降(92.75% → 77.76%),WOE 单调上升(-0.3898 → 0.9071),符合风控分箱预期。IV=0.2452,该变量区分度尚可。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 1 箱((-inf, 1.5])为极高风险区间:坏率高达 92.75%,策略上应重点管控

acc_ps_perd_no
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 2.5]304444.39%205283993.27%0.07
2(2.5, 5.5]181270.81%197161589.13%0.12
3(5.5, 6.5]35175.93%2332893.45%0.07
4(6.5, 16.5]115592.77%18397284.16%0.19
5(16.5, inf]496100.00%10239479.44%0.26
合计6858710614889.65%0.12
2026-11-03T11:57:52.809609 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(93.27% → 93.45% → 79.44%),如需严格单调可考虑调整切点或合并分箱。IV=0.2055,该变量区分度尚可。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 89.13% 回升至 93.45%,如需严格单调可考虑合并这两箱

第 3 箱((5.5, 6.5])为极高风险区间:坏率高达 93.45%,策略上应重点管控

tothismonthodcontiaddnumm12
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]299243.63%445254785.13%0.17
2(0.5, inf]3866100.00%265360193.15%0.07
合计6858710614889.65%0.12
2026-11-03T11:57:52.926797 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调上升(85.13% → 93.15%),WOE 单调下降(0.4140 → -0.4506),符合风控分箱预期。IV=0.1837,该变量区分度尚可。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 1 箱 WOE<0(风险高于平均)

第 2 箱((0.5, inf])为极高风险区间:坏率高达 93.15%,策略上应重点管控

loan_os_prcp_orig_prcp_pc
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 13.04]5508.02%11543579.09%0.26
2(13.04, 53.4]334156.74%372296988.87%0.13
3(53.4, 67.97]106072.19%9796390.85%0.10
4(67.97, 76.46]55980.34%2853194.99%0.05
5(76.46, inf]1348100.00%98125092.73%0.08
合计6858710614889.65%0.12
2026-11-03T11:57:53.050319 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(79.09% → 94.99% → 92.73%),如需严格单调可考虑调整切点或合并分箱。IV=0.1437,该变量区分度尚可。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 4→5 箱有局部反弹:Bad Rate 由 94.99% 回落至 92.73%,如需严格单调可考虑合并这两箱

第 4 箱((67.97, 76.46])为极高风险区间:坏率高达 94.99%,策略上应重点管控

all_repay_amount_m3_shd_amt_pc
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 66.67]313645.73%229290792.70%0.08
2(66.67, 66.73]37751.22%3833989.92%0.11
3(66.73, 69.51]61760.22%8453386.39%0.16
4(69.51, 99.99]69070.28%13056081.16%0.23
5(99.99, inf]2038100.00%229180988.76%0.13
合计6858710614889.65%0.12
2026-11-03T11:57:53.242992 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(92.70% → 92.70% → 88.76%),如需严格单调可考虑调整切点或合并分箱。IV=0.1342,该变量区分度尚可。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 4→5 箱有局部反弹:Bad Rate 由 81.16% 回升至 88.76%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 66.67])为极高风险区间:坏率高达 92.70%,策略上应重点管控

ovd_days_d30_num_m6
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]522176.13%437478491.63%0.09
2(0.5, 1.5]112892.58%17695284.40%0.18
3(1.5, inf]509100.00%9741280.94%0.24
合计6858710614889.65%0.12
2026-11-03T11:57:53.355118 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调下降(91.63% → 80.94%),WOE 单调上升(-0.2345 → 0.7123),符合风控分箱预期。IV=0.1315,该变量区分度尚可。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 1 箱((-inf, 0.5])为极高风险区间:坏率高达 91.63%,策略上应重点管控

od_amt_m3_loan_os_prcp_m3_pc
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 10.09]490371.49%415448891.54%0.09
2(10.09, 14.35]71681.93%8063688.83%0.13
3(14.35, 24.15]54389.85%9345082.87%0.21
4(24.15, 25.53]35495.01%4530987.29%0.15
5(25.53, inf]342100.00%7726577.49%0.29
合计6858710614889.65%0.12
2026-11-03T11:57:53.472861 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(91.54% → 91.54% → 77.49%),如需严格单调可考虑调整切点或合并分箱。IV=0.1301,该变量区分度尚可。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 82.87% 回升至 87.29%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 10.09])为极高风险区间:坏率高达 91.54%,策略上应重点管控

repay_amount_last_month_loan_os_prcp_pc
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.79]455866.46%382417691.62%0.09
2(0.79, 8.66]34471.48%5029485.47%0.17
3(8.66, 20.06]81983.42%8073990.23%0.11
4(20.06, 49.49]60292.20%8851485.38%0.17
5(49.49, inf]535100.00%11042579.44%0.26
合计6858710614889.65%0.12
2026-11-03T11:57:53.599348 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(91.62% → 91.62% → 79.44%),如需严格单调可考虑调整切点或合并分箱。IV=0.1273,该变量区分度尚可。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 85.47% 回升至 90.23%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 0.79])为极高风险区间:坏率高达 91.62%,策略上应重点管控

max_od_days_m12
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 4.5]4546.62%2842693.83%0.07
2(4.5, 7.5]53514.42%6646987.66%0.14
3(7.5, 9.5]40820.37%2238694.61%0.06
4(9.5, 29.5]457987.14%433414690.54%0.10
5(29.5, inf]882100.00%16172181.75%0.22
合计6858710614889.65%0.12
2026-11-03T11:57:53.725677 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(93.83% → 94.61% → 81.75%),如需严格单调可考虑调整切点或合并分箱。IV=0.1213,该变量区分度尚可。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 87.66% 回升至 94.61%,如需严格单调可考虑合并这两箱

第 3 箱((7.5, 9.5])为极高风险区间:坏率高达 94.61%,策略上应重点管控

ps_od_norm_int_amt
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 10.25]4817.01%8939281.50%0.23
2(10.25, 44.34]148928.73%162132789.12%0.12
3(44.34, 107.69]155351.37%195135887.44%0.14
4(107.69, 757.16]282992.62%243258691.41%0.09
5(757.16, inf]506100.00%2148595.85%0.04
合计6858710614889.65%0.12
2026-11-03T11:57:53.854985 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(81.50% → 95.85% → 95.85%),如需严格单调可考虑调整切点或合并分箱。IV=0.1186,该变量区分度尚可。

WOE 在第 3、4 箱之间穿越 0 线:前 3 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 89.12% 回落至 87.44%,如需严格单调可考虑合并这两箱

第 5 箱((757.16, inf])为极高风险区间:坏率高达 95.85%,策略上应重点管控

repay_amount_all_shd_amt_pc
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 33.16]453566.13%379415691.64%0.09
2(33.16, 84.39]36671.46%5531184.97%0.18
3(84.39, 90.67]77082.69%9867287.27%0.15
4(90.67, 96.6]67692.55%7859888.46%0.13
5(96.6, inf]511100.00%10041180.43%0.24
合计6858710614889.65%0.12
2026-11-03T11:57:53.981076 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(91.64% → 91.64% → 80.43%),如需严格单调可考虑调整切点或合并分箱。IV=0.1084,该变量区分度尚可。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 84.97% 回升至 87.27%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 33.16])为极高风险区间:坏率高达 91.64%,策略上应重点管控

end_loan_to_cur_freq
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, -362.5]4206.12%8433680.00%0.25
2(-362.5, -248.5]176631.88%199156788.73%0.13
3(-248.5, -109.5]187759.24%159171891.53%0.09
4(-109.5, -26.5]56267.44%2853495.02%0.05
5(-26.5, inf]2233100.00%240199389.25%0.12
合计6858710614889.65%0.12
2026-11-03T11:57:54.111168 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(80.00% → 95.02% → 89.25%),如需严格单调可考虑调整切点或合并分箱。IV=0.1019,该变量区分度尚可。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 4→5 箱有局部反弹:Bad Rate 由 95.02% 回落至 89.25%,如需严格单调可考虑合并这两箱

第 4 箱((-109.5, -26.5])为极高风险区间:坏率高达 95.02%,策略上应重点管控

cur_od_loan_amt
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 124.01]5728.34%8648684.97%0.18
2(124.01, 558.05]195836.89%227173188.41%0.13
3(558.05, 725.59]44543.38%6837784.72%0.18
4(725.59, 3545.66]301287.30%283272990.60%0.10
5(3545.66, inf]871100.00%4682594.72%0.06
合计6858710614889.65%0.12
2026-11-03T11:57:54.236696 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(84.97% → 94.72% → 94.72%),如需严格单调可考虑调整切点或合并分箱。IV=0.0936,该变量区分度偏弱但可用。

WOE 在第 3、4 箱之间穿越 0 线:前 3 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 88.41% 回落至 84.72%,如需严格单调可考虑合并这两箱

第 5 箱((3545.66, inf])为极高风险区间:坏率高达 94.72%,策略上应重点管控

odduecontireducenumm12
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]510274.39%440466291.38%0.09
2(0.5, inf]1756100.00%270148684.62%0.18
合计6858710614889.65%0.12
2026-11-03T11:57:54.351861 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调下降(91.38% → 84.62%),WOE 单调上升(-0.2018 → 0.4532),符合风控分箱预期。IV=0.0908,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 1 箱 WOE<0(风险高于平均)

第 1 箱((-inf, 0.5])为极高风险区间:坏率高达 91.38%,策略上应重点管控

max_setl_rem_prcp_paym_m3_last_month_loan_os_prcp_pc
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 1.88]479569.92%544425188.65%0.13
2(1.88, 11.68]54477.85%2052496.32%0.04
3(11.68, 20.03]66587.55%5860791.28%0.10
4(20.03, 33.46]35892.77%2433493.30%0.07
5(33.46, inf]496100.00%6443287.10%0.15
合计6858710614889.65%0.12
2026-11-03T11:57:54.466627 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.65% → 96.32% → 87.10%),如需严格单调可考虑调整切点或合并分箱。IV=0.0888,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 88.65% 回升至 96.32%,如需严格单调可考虑合并这两箱

第 2 箱((1.88, 11.68])为极高风险区间:坏率高达 96.32%,策略上应重点管控

apply_amt_avg_m12
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 4525]300443.80%381262387.32%0.15
2(4525, 10810]116760.82%124104389.37%0.12
3(10810, 32366.67]148182.41%110137192.57%0.08
4(32366.67, 48025]50289.73%5344989.44%0.12
5(48025, inf]704100.00%4266294.03%0.06
合计6858710614889.65%0.12
2026-11-03T11:57:54.661756 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.32% → 94.03% → 94.03%),如需严格单调可考虑调整切点或合并分箱。IV=0.0793,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 92.57% 回落至 89.44%,如需严格单调可考虑合并这两箱

第 5 箱((48025, inf])为极高风险区间:坏率高达 94.03%,策略上应重点管控

orig_prcp_avg_m6
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 636.67]460667.16%542406488.23%0.13
2(636.67, 2158.33]47074.02%4642490.21%0.11
3(2158.33, 5016.67]52781.70%4748091.08%0.10
4(5016.67, 15012.5]63190.90%4658592.71%0.08
5(15012.5, inf]624100.00%2959595.35%0.05
合计6858710614889.65%0.12
2026-11-03T11:57:54.786418 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调上升(88.23% → 95.35%),WOE 单调下降(0.1440 → -0.8626),符合风控分箱预期。IV=0.0769,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 5 箱((15012.5, inf])为极高风险区间:坏率高达 95.35%,策略上应重点管控

account_age_loan_day
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.52]164523.99%118152792.83%0.08
2(0.52, 0.96]219956.05%240195989.09%0.12
3(0.96, 1.19]64665.47%10753983.44%0.20
4(1.19, 4.62]145386.66%137131690.57%0.10
5(4.62, inf]915100.00%10880788.20%0.13
合计6858710614889.65%0.12
2026-11-03T11:57:54.905428 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(92.83% → 92.83% → 88.20%),如需严格单调可考虑调整切点或合并分箱。IV=0.0736,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 83.44% 回升至 90.57%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 0.52])为极高风险区间:坏率高达 92.83%,策略上应重点管控

repay_amount_m6_loan_os_prcp_pc
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 42.01]219431.99%173202192.11%0.09
2(42.01, 70.26]229865.50%228207090.08%0.11
3(70.26, 82.59]52373.13%6945486.81%0.15
4(82.59, 99.67]34978.22%6828180.52%0.24
5(99.67, inf]1494100.00%172132288.49%0.13
合计6858710614889.65%0.12
2026-11-03T11:57:55.024185 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(92.11% → 92.11% → 88.49%),如需严格单调可考虑调整切点或合并分箱。IV=0.0729,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 4→5 箱有局部反弹:Bad Rate 由 80.52% 回升至 88.49%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 42.01])为极高风险区间:坏率高达 92.11%,策略上应重点管控

sum_orig_prcp
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 10450]250536.53%314219187.47%0.14
2(10450, 12554]35541.70%2133494.08%0.06
3(12554, 19150]82553.73%9672988.36%0.13
4(19150, 110100]280994.69%263254690.64%0.10
5(110100, inf]364100.00%1634895.60%0.05
合计6858710614889.65%0.12
2026-11-03T11:57:55.151396 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.47% → 95.60% → 95.60%),如需严格单调可考虑调整切点或合并分箱。IV=0.0720,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 94.08% 回落至 88.36%,如需严格单调可考虑合并这两箱

第 5 箱((110100, inf])为极高风险区间:坏率高达 95.60%,策略上应重点管控

loannum
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 9.5]4356.74%7036583.91%0.19
2(9.5, 27.5]202038.03%242177888.02%0.14
3(27.5, 39.5]103854.11%8795191.62%0.09
4(39.5, 61.5]121272.89%137107588.70%0.13
5(61.5, inf]1750100.00%130162092.57%0.08
合计6455666578989.68%0.12
2026-11-03T11:57:55.272907 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(83.91% → 92.57% → 92.57%),如需严格单调可考虑调整切点或合并分箱。IV=0.0712,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 91.62% 回落至 88.70%,如需严格单调可考虑合并这两箱

第 5 箱((61.5, inf])为极高风险区间:坏率高达 92.57%,策略上应重点管控

sum_card_mob12_neverdur
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]209832.50%252184687.99%0.14
2(0.5, 1.5]83445.42%7775790.77%0.10
3(1.5, 6.5]185774.19%214164388.48%0.13
4(6.5, 8.5]33579.38%1332296.12%0.04
5(8.5, inf]1331100.00%110122191.74%0.09
合计6455666578989.68%0.12
2026-11-03T11:57:55.398610 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.99% → 96.12% → 91.74%),如需严格单调可考虑调整切点或合并分箱。IV=0.0657,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 90.77% 回落至 88.48%,如需严格单调可考虑合并这两箱

第 4 箱((6.5, 8.5])为极高风险区间:坏率高达 96.12%,策略上应重点管控

od_amt_m12_loan_os_prcp_m12_pc
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.91]124118.10%129111289.61%0.12
2(0.91, 1.09]38523.71%5133486.75%0.15
3(1.09, 2.17]354075.33%299324191.55%0.09
4(2.17, 3.17]78286.73%8969388.62%0.13
5(3.17, inf]910100.00%14276884.40%0.18
合计6858710614889.65%0.12
2026-11-03T11:57:55.529122 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(89.61% → 91.55% → 84.40%),如需严格单调可考虑调整切点或合并分箱。IV=0.0654,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 86.75% 回升至 91.55%,如需严格单调可考虑合并这两箱

第 3 箱((1.09, 2.17])为极高风险区间:坏率高达 91.55%,策略上应重点管控

setl_int_mon
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 3.64]500973.04%457455290.88%0.10
2(3.64, 17.71]34478.05%5928582.85%0.21
3(17.71, 42.33]37083.45%4632487.57%0.14
4(42.33, 90.6]37788.95%6731082.23%0.22
5(90.6, inf]758100.00%8167789.31%0.12
合计6858710614889.65%0.12
2026-11-03T11:57:55.657383 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(90.88% → 90.88% → 89.31%),如需严格单调可考虑调整切点或合并分箱。IV=0.0652,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 82.85% 回升至 87.57%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 3.64])为极高风险区间:坏率高达 90.88%,策略上应重点管控

setl_rem_prcp_paym_m3_loan_os_prcp_pc
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, -4999]3475.06%2232593.66%0.07
2(-4999, 3.21]450670.76%525398188.35%0.13
3(3.21, 19.05]85783.26%5080794.17%0.06
4(19.05, 25.49]36588.58%4232388.49%0.13
5(25.49, inf]783100.00%7171290.93%0.10
合计6858710614889.65%0.12
2026-11-03T11:57:55.783734 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(93.66% → 94.17% → 90.93%),如需严格单调可考虑调整切点或合并分箱。IV=0.0649,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 88.35% 回升至 94.17%,如需严格单调可考虑合并这两箱

第 3 箱((3.21, 19.05])为极高风险区间:坏率高达 94.17%,策略上应重点管控

sum_month3_new_loan
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]141621.94%182123487.15%0.15
2(0.5, 1.5]132842.51%142118689.31%0.12
3(1.5, 2.5]108859.36%13695287.50%0.14
4(2.5, 3.5]75671.08%5270493.12%0.07
5(3.5, inf]1867100.00%154171391.75%0.09
合计6455666578989.68%0.12
2026-11-03T11:57:55.913778 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.15% → 93.12% → 91.75%),如需严格单调可考虑调整切点或合并分箱。IV=0.0591,该变量区分度偏弱但可用。

WOE 在第 3、4 箱之间穿越 0 线:前 3 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 89.31% 回落至 87.50%,如需严格单调可考虑合并这两箱

第 4 箱((2.5, 3.5])为极高风险区间:坏率高达 93.12%,策略上应重点管控

sum_apply_amt
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 2050]5137.48%7543885.38%0.17
2(2050, 9625]177633.38%187158989.47%0.12
3(9625, 11585]39039.06%6332783.85%0.19
4(11585, 130400]381794.72%366345190.41%0.11
5(130400, inf]362100.00%1934394.75%0.06
合计6858710614889.65%0.12
2026-11-03T11:57:56.114678 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(85.38% → 94.75% → 94.75%),如需严格单调可考虑调整切点或合并分箱。IV=0.0571,该变量区分度偏弱但可用。

WOE 在第 3、4 箱之间穿越 0 线:前 3 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 89.47% 回落至 83.85%,如需严格单调可考虑合并这两箱

第 5 箱((130400, inf])为极高风险区间:坏率高达 94.75%,策略上应重点管控

loan_reject_rate_3m
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, -0.15]82512.82%6576092.12%0.09
2(-0.15, 0.02]152036.44%172134888.68%0.13
3(0.02, 0.28]46343.64%2843593.95%0.06
4(0.28, 0.58]134164.48%115122691.42%0.09
5(0.58, inf]2286100.00%284200287.58%0.14
合计6435664577189.68%0.12
2026-11-03T11:57:56.243753 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(92.12% → 93.95% → 87.58%),如需严格单调可考虑调整切点或合并分箱。IV=0.0569,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 88.68% 回升至 93.95%,如需严格单调可考虑合并这两箱

第 3 箱((0.02, 0.28])为极高风险区间:坏率高达 93.95%,策略上应重点管控

setl_prcp_mon_ratio
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0]499072.76%458453290.82%0.10
2(0, 0.01]50980.18%6944086.44%0.16
3(0.01, 0.02]37385.62%3334091.15%0.10
4(0.02, 0.04]47292.51%7639683.90%0.19
5(0.04, inf]514100.00%7444085.60%0.17
合计6858710614889.65%0.12
2026-11-03T11:57:56.377911 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(90.82% → 91.15% → 85.60%),如需严格单调可考虑调整切点或合并分箱。IV=0.0556,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 86.44% 回升至 91.15%,如需严格单调可考虑合并这两箱

第 3 箱((0.01, 0.02])为极高风险区间:坏率高达 91.15%,策略上应重点管控

useratemore90_m6_num
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]440664.25%515389188.31%0.13
2(0.5, 1.5]57072.56%5351790.70%0.10
3(1.5, 2.5]86085.10%5380793.84%0.07
4(2.5, 3.5]44191.53%3141092.97%0.08
5(3.5, inf]581100.00%5852390.02%0.11
合计6858710614889.65%0.12
2026-11-03T11:57:56.506953 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.31% → 93.84% → 90.02%),如需严格单调可考虑调整切点或合并分箱。IV=0.0556,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 93.84% 回落至 92.97%,如需严格单调可考虑合并这两箱

第 3 箱((1.5, 2.5])为极高风险区间:坏率高达 93.84%,策略上应重点管控

account_age
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 6.5]189727.66%151174692.04%0.09
2(6.5, 11.5]251464.32%270224489.26%0.12
3(11.5, 14.5]71774.77%11360484.24%0.19
4(14.5, 33.5]135094.46%146120489.19%0.12
5(33.5, inf]380100.00%3035092.11%0.09
合计6858710614889.65%0.12
2026-11-03T11:57:56.667511 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(92.04% → 92.11% → 92.11%),如需严格单调可考虑调整切点或合并分箱。IV=0.0555,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 92.04% 回落至 89.26%,如需严格单调可考虑合并这两箱

第 5 箱((33.5, inf])为极高风险区间:坏率高达 92.11%,策略上应重点管控

apply_num_m6
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]448065.33%531394988.15%0.13
2(0.5, 1.5]143986.31%112132792.22%0.08
3(1.5, 2.5]46493.07%3742792.03%0.09
4(2.5, inf]475100.00%3044593.68%0.07
合计6858710614889.65%0.12
2026-11-03T11:57:56.794713 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.15% → 93.68% → 93.68%),如需严格单调可考虑调整切点或合并分箱。IV=0.0555,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 92.22% 回落至 92.03%,如需严格单调可考虑合并这两箱

第 4 箱((2.5, inf])为极高风险区间:坏率高达 93.68%,策略上应重点管控

repay_amount_m12_all_shd_amt_pc
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 85.33]118317.25%111107290.62%0.10
2(85.33, 86.76]37422.70%2235294.12%0.06
3(86.76, 89.01]47929.69%8139883.09%0.20
4(89.01, 99.76]447894.98%471400789.48%0.12
5(99.76, inf]344100.00%2531992.73%0.08
合计6858710614889.65%0.12
2026-11-03T11:57:56.918109 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(90.62% → 94.12% → 92.73%),如需严格单调可考虑调整切点或合并分箱。IV=0.0527,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 94.12% 回落至 83.09%,如需严格单调可考虑合并这两箱

第 2 箱((85.33, 86.76])为极高风险区间:坏率高达 94.12%,策略上应重点管控

t_ps_od_days_max
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]503673.43%466457090.75%0.10
2(0.5, 3.5]77184.67%10966285.86%0.16
3(3.5, 11.5]43290.97%7335983.10%0.20
4(11.5, inf]619100.00%6255789.98%0.11
合计6858710614889.65%0.12
2026-11-03T11:57:57.048366 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(90.75% → 90.75% → 89.98%),如需严格单调可考虑调整切点或合并分箱。IV=0.0523,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 83.10% 回升至 89.98%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 0.5])为极高风险区间:坏率高达 90.75%,策略上应重点管控

nosvaing_bank_cnt_1m
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]428865.57%492379688.53%0.13
2(0.5, 1.5]142187.29%135128690.50%0.10
3(1.5, inf]831100.00%4878394.22%0.06
合计6540675586589.68%0.12
2026-11-03T11:57:57.178063 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调上升(88.53% → 94.22%),WOE 单调下降(0.1188 → -0.6299),符合风控分箱预期。IV=0.0508,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 3 箱((1.5, inf])为极高风险区间:坏率高达 94.22%,策略上应重点管控

last6average
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 128.5]149322.80%158133589.42%0.12
2(128.5, 977.5]128942.48%166112387.12%0.15
3(977.5, 2255.5]116260.22%113104990.28%0.11
4(2255.5, 3242.5]64970.13%8756286.59%0.15
5(3242.5, inf]1956100.00%152180492.23%0.08
合计6549676587389.68%0.12
2026-11-03T11:57:57.300377 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(89.42% → 92.23% → 92.23%),如需严格单调可考虑调整切点或合并分箱。IV=0.0500,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 89.42% 回落至 87.12%,如需严格单调可考虑合并这两箱

第 5 箱((3242.5, inf])为极高风险区间:坏率高达 92.23%,策略上应重点管控

useratemore90monthsubcurmonth
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 1.5]89313.02%7681791.49%0.09
2(1.5, 3.5]76324.15%4771693.84%0.07
3(3.5, 5.5]79635.75%7272490.95%0.10
4(5.5, 8.5]97649.99%12085687.70%0.14
5(8.5, inf]3430100.00%395303588.48%0.13
合计6858710614889.65%0.12
2026-11-03T11:57:57.431135 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(91.49% → 93.84% → 88.48%),如需严格单调可考虑调整切点或合并分箱。IV=0.0497,该变量区分度偏弱但可用。

WOE 在第 3、4 箱之间穿越 0 线:前 3 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 91.49% 回升至 93.84%,如需严格单调可考虑合并这两箱

第 2 箱((1.5, 3.5])为极高风险区间:坏率高达 93.84%,策略上应重点管控

setl_int_avg
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 13.19]102714.98%12490387.93%0.14
2(13.19, 16.74]34219.96%5328984.50%0.18
3(16.74, 22.2]47226.84%3543792.58%0.08
4(22.2, 110.95]328774.77%363292488.96%0.12
5(110.95, inf]1730100.00%135159592.20%0.08
合计6858710614889.65%0.12
2026-11-03T11:57:57.636274 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.93% → 92.58% → 92.20%),如需严格单调可考虑调整切点或合并分箱。IV=0.0497,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 87.93% 回落至 84.50%,如需严格单调可考虑合并这两箱

第 3 箱((16.74, 22.2])为极高风险区间:坏率高达 92.58%,策略上应重点管控

bank_997_cnt_6m
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 1.5]326949.98%391287888.04%0.14
2(1.5, 3.5]158074.14%129145191.84%0.09
3(3.5, 5.5]78386.12%8270189.53%0.12
4(5.5, 9.5]57594.91%3753893.57%0.07
5(9.5, inf]333100.00%3629789.19%0.12
合计6540675586589.68%0.12
2026-11-03T11:57:57.760890 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.04% → 93.57% → 89.19%),如需严格单调可考虑调整切点或合并分箱。IV=0.0484,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 91.84% 回落至 89.53%,如需严格单调可考虑合并这两箱

第 4 箱((5.5, 9.5])为极高风险区间:坏率高达 93.57%,策略上应重点管控

total_998_rate_3m
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, -4999]5628.59%7648686.48%0.16
2(-4999, 0.1]94523.04%9784889.74%0.11
3(0.1, 0.52]199753.58%158183992.09%0.09
4(0.52, 0.73]120672.02%115109190.46%0.11
5(0.73, inf]1830100.00%229160187.49%0.14
合计6540675586589.68%0.12
2026-11-03T11:57:57.885332 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(86.48% → 92.09% → 87.49%),如需严格单调可考虑调整切点或合并分箱。IV=0.0481,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 92.09% 回落至 90.46%,如需严格单调可考虑合并这两箱

第 3 箱((0.1, 0.52])为极高风险区间:坏率高达 92.09%,策略上应重点管控

ovd_days_d15_num_m3
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]228633.33%177210992.26%0.08
2(0.5, 1.5]276973.71%309246088.84%0.13
3(1.5, 2.5]87886.51%10777187.81%0.14
4(2.5, 4.5]57094.82%7949186.14%0.16
5(4.5, inf]355100.00%3831789.30%0.12
合计6858710614889.65%0.12
2026-11-03T11:57:58.011806 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(92.26% → 92.26% → 89.30%),如需严格单调可考虑调整切点或合并分箱。IV=0.0480,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 4→5 箱有局部反弹:Bad Rate 由 86.14% 回升至 89.30%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 0.5])为极高风险区间:坏率高达 92.26%,策略上应重点管控

loan_rate_1_12
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.02]251138.39%298221388.13%0.13
2(0.02, 0.05]62347.92%6356089.89%0.11
3(0.05, 0.07]44054.65%2341794.77%0.06
4(0.07, 0.27]237190.90%247212489.58%0.12
5(0.27, inf]595100.00%4455192.61%0.08
合计6540675586589.68%0.12
2026-11-03T11:57:58.138255 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.13% → 94.77% → 92.61%),如需严格单调可考虑调整切点或合并分箱。IV=0.0479,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 94.77% 回落至 89.58%,如需严格单调可考虑合并这两箱

第 3 箱((0.05, 0.07])为极高风险区间:坏率高达 94.77%,策略上应重点管控

nosvaing_bank_rate_1_3
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, -4999]205031.35%253179787.66%0.14
2(-4999, 0.28]258070.80%278230289.22%0.12
3(0.28, 0.39]33975.98%2731292.04%0.09
4(0.39, 0.74]79988.20%5574493.12%0.07
5(0.74, inf]772100.00%6271091.97%0.09
合计6540675586589.68%0.12
2026-11-03T11:57:58.263551 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.66% → 93.12% → 91.97%),如需严格单调可考虑调整切点或合并分箱。IV=0.0466,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 4→5 箱有局部反弹:Bad Rate 由 93.12% 回落至 91.97%,如需严格单调可考虑合并这两箱

第 4 箱((0.39, 0.74])为极高风险区间:坏率高达 93.12%,策略上应重点管控

allcardsoverdraftrate6avg
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.07]314448.71%368277688.30%0.13
2(0.07, 0.88]189678.08%154174291.88%0.09
3(0.88, 0.95]42884.71%5637286.92%0.15
4(0.95, 0.99]38090.60%2535593.42%0.07
5(0.99, inf]607100.00%6354489.62%0.12
合计6455666578989.68%0.12
2026-11-03T11:57:58.394445 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.30% → 93.42% → 89.62%),如需严格单调可考虑调整切点或合并分箱。IV=0.0457,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 91.88% 回落至 86.92%,如需严格单调可考虑合并这两箱

第 4 箱((0.95, 0.99])为极高风险区间:坏率高达 93.42%,策略上应重点管控

f_ps_od_days_max
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]577384.18%548522590.51%0.10
2(0.5, 2.5]64293.54%10753583.33%0.20
3(2.5, inf]443100.00%5538887.58%0.14
合计6858710614889.65%0.12
2026-11-03T11:57:58.521740 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(90.51% → 90.51% → 87.58%),如需严格单调可考虑调整切点或合并分箱。IV=0.0454,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 83.33% 回升至 87.58%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 0.5])为极高风险区间:坏率高达 90.51%,策略上应重点管控

setl_rem_prcp_paym_m12_loan_os_prcp_pc
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, -9998.5]442964.58%425400490.40%0.11
2(-9998.5, -4999]55972.73%4151892.67%0.08
3(-4999, 17.22]80584.47%10570086.96%0.15
4(17.22, 79.79]63693.74%9753984.75%0.18
5(79.79, inf]429100.00%4238790.21%0.11
合计6858710614889.65%0.12
2026-11-03T11:57:58.647547 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(90.40% → 92.67% → 90.21%),如需严格单调可考虑调整切点或合并分箱。IV=0.0453,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 90.40% 回升至 92.67%,如需严格单调可考虑合并这两箱

第 2 箱((-9998.5, -4999])为极高风险区间:坏率高达 92.67%,策略上应重点管控

cardaveragemob
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 20.5]214233.18%259188387.91%0.14
2(20.5, 75.5]183361.58%171166290.67%0.10
3(75.5, 84.5]48269.05%6242087.14%0.15
4(84.5, 104.5]87782.63%5981893.27%0.07
5(104.5, inf]1121100.00%115100689.74%0.11
合计6455666578989.68%0.12
2026-11-03T11:57:58.778594 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.91% → 93.27% → 89.74%),如需严格单调可考虑调整切点或合并分箱。IV=0.0445,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 90.67% 回落至 87.14%,如需严格单调可考虑合并这两箱

第 4 箱((84.5, 104.5])为极高风险区间:坏率高达 93.27%,策略上应重点管控

sum_open_loan_mob24_neverdur
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]71011.00%9461686.76%0.15
2(0.5, 3.5]311959.31%357276288.55%0.13
3(3.5, 4.5]74070.77%6967190.68%0.10
4(4.5, 8.5]147793.65%117136092.08%0.09
5(8.5, inf]410100.00%2938192.93%0.08
合计6456666579089.68%0.12
2026-11-03T11:57:58.910513 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调上升(86.76% → 92.93%),WOE 单调下降(0.2826 → -0.4129),符合风控分箱预期。IV=0.0445,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 5 箱((8.5, inf])为极高风险区间:坏率高达 92.93%,策略上应重点管控

ncarloanbalancerate6month
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.19]153123.78%181135088.18%0.13
2(0.19, 0.63]34429.12%5129385.17%0.17
3(0.63, 2.04]344682.64%326312090.54%0.10
4(2.04, 2.81]36988.37%2034994.58%0.06
5(2.81, inf]749100.00%8666388.52%0.13
合计6439664577589.69%0.11
2026-11-03T11:57:59.110028 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.18% → 94.58% → 88.52%),如需严格单调可考虑调整切点或合并分箱。IV=0.0444,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 88.18% 回落至 85.17%,如需严格单调可考虑合并这两箱

第 4 箱((2.04, 2.81])为极高风险区间:坏率高达 94.58%,策略上应重点管控

nosvaing_bank_rate_1m
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, -4999]242437.06%287213788.16%0.13
2(-4999, 0.04]186465.57%205165989.00%0.12
3(0.04, 0.35]69376.16%7062389.90%0.11
4(0.35, 0.61]76087.78%6070092.11%0.09
5(0.61, inf]799100.00%5374693.37%0.07
合计6540675586589.68%0.12
2026-11-03T11:57:59.233554 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调上升(88.16% → 93.37%),WOE 单调下降(0.1544 → -0.4824),符合风控分箱预期。IV=0.0434,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 5 箱((0.61, inf])为极高风险区间:坏率高达 93.37%,策略上应重点管控

setl_prcp_avg
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 88.54]72010.50%10062086.11%0.16
2(88.54, 609.66]347561.17%375310089.21%0.12
3(609.66, 719.26]35866.39%2833092.18%0.08
4(719.26, 936.76]49573.61%6343287.27%0.15
5(936.76, inf]1810100.00%144166692.04%0.09
合计6858710614889.65%0.12
2026-11-03T11:57:59.362095 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(86.11% → 92.18% → 92.04%),如需严格单调可考虑调整切点或合并分箱。IV=0.0429,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 92.18% 回落至 87.27%,如需严格单调可考虑合并这两箱

第 3 箱((609.66, 719.26])为极高风险区间:坏率高达 92.18%,策略上应重点管控

bank_total_used_limit
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.16]112717.23%14598287.13%0.15
2(0.16, 0.21]30621.91%2528191.83%0.09
3(0.21, 0.31]76433.59%8567988.87%0.13
4(0.31, 0.75]301479.68%257275791.47%0.09
5(0.75, inf]1329100.00%163116687.74%0.14
合计6540675586589.68%0.12
2026-11-03T11:57:59.484712 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.13% → 91.83% → 87.74%),如需严格单调可考虑调整切点或合并分箱。IV=0.0426,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 91.83% 回落至 88.87%,如需严格单调可考虑合并这两箱

第 2 箱((0.16, 0.21])为极高风险区间:坏率高达 91.83%,策略上应重点管控

rh_cr_bal_pro
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, -4999]3926.09%3435891.33%0.09
2(-4999, 0.45]112423.55%15097486.65%0.15
3(0.45, 0.82]301170.32%263274891.27%0.10
4(0.82, 0.97]147393.20%157131689.34%0.12
5(0.97, inf]438100.00%6037886.30%0.16
合计6438664577489.69%0.11
2026-11-03T11:57:59.605205 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(91.33% → 91.33% → 86.30%),如需严格单调可考虑调整切点或合并分箱。IV=0.0418,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 86.65% 回升至 91.27%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, -4999])为极高风险区间:坏率高达 91.33%,策略上应重点管控

loan_reject_rate_6m
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.25]300246.65%272273090.94%0.10
2(0.25, 0.51]156670.99%181138588.44%0.13
3(0.51, 0.62]47478.35%3444092.83%0.08
4(0.62, 0.69]44385.24%4739689.39%0.12
5(0.69, inf]950100.00%13082086.32%0.16
合计6435664577189.68%0.12
2026-11-03T11:57:59.733382 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(90.94% → 92.83% → 86.32%),如需严格单调可考虑调整切点或合并分箱。IV=0.0405,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 88.44% 回升至 92.83%,如需严格单调可考虑合并这两箱

第 3 箱((0.51, 0.62])为极高风险区间:坏率高达 92.83%,策略上应重点管控

no_bank_998_rate_12m
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.22]110116.83%95100691.37%0.09
2(0.22, 0.33]65126.79%4560693.09%0.07
3(0.33, 0.68]383385.40%432340188.73%0.13
4(0.68, 0.74]34590.67%5029585.51%0.17
5(0.74, inf]610100.00%5355791.31%0.10
合计6540675586589.68%0.12
2026-11-03T11:57:59.862259 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(91.37% → 93.09% → 91.31%),如需严格单调可考虑调整切点或合并分箱。IV=0.0404,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 91.37% 回升至 93.09%,如需严格单调可考虑合并这两箱

第 2 箱((0.22, 0.33])为极高风险区间:坏率高达 93.09%,策略上应重点管控

max_setl_rem_prcp_paym_m6_last_month_loan_os_prcp_pc
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 3.39]339249.46%383300988.71%0.13
2(3.39, 9.1]51456.96%3947592.41%0.08
3(9.1, 11.68]51964.52%2949094.41%0.06
4(11.68, 90.9]180890.89%199160988.99%0.12
5(90.9, inf]625100.00%6056590.40%0.11
合计6858710614889.65%0.12
2026-11-03T11:57:59.991022 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.71% → 94.41% → 90.40%),如需严格单调可考虑调整切点或合并分箱。IV=0.0404,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 94.41% 回落至 88.99%,如需严格单调可考虑合并这两箱

第 3 箱((9.1, 11.68])为极高风险区间:坏率高达 94.41%,策略上应重点管控

c_bank_rate_6_12
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.24]135320.69%159119488.25%0.13
2(0.24, 0.39]84233.56%8975389.43%0.12
3(0.39, 0.44]38639.46%1936795.08%0.05
4(0.44, 0.68]207571.19%194188190.65%0.10
5(0.68, inf]1884100.00%214167088.64%0.13
合计6540675586589.68%0.12
2026-11-03T11:58:00.119916 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.25% → 95.08% → 88.64%),如需严格单调可考虑调整切点或合并分箱。IV=0.0394,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 95.08% 回落至 90.65%,如需严格单调可考虑合并这两箱

第 3 箱((0.39, 0.44])为极高风险区间:坏率高达 95.08%,策略上应重点管控

no_bank_total_used_limit
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.27]348253.24%396308688.63%0.13
2(0.27, 0.3]38059.05%2335793.95%0.06
3(0.3, 0.33]37564.79%4333288.53%0.13
4(0.33, 0.58]169090.63%140155091.72%0.09
5(0.58, inf]613100.00%7354088.09%0.14
合计6540675586589.68%0.12
2026-11-03T11:58:00.248719 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.63% → 93.95% → 88.09%),如需严格单调可考虑调整切点或合并分箱。IV=0.0393,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 88.63% 回升至 93.95%,如需严格单调可考虑合并这两箱

第 2 箱((0.27, 0.3])为极高风险区间:坏率高达 93.95%,策略上应重点管控

mostcarduserate
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.15]320749.68%369283888.49%0.13
2(0.15, 0.82]69360.42%6562890.62%0.10
3(0.82, 0.99]78672.59%5373393.26%0.07
4(0.99, 1.07]143394.79%134129990.65%0.10
5(1.07, inf]336100.00%4529186.61%0.15
合计6455666578989.68%0.12
2026-11-03T11:58:00.384120 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.49% → 93.26% → 86.61%),如需严格单调可考虑调整切点或合并分箱。IV=0.0385,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 88.49% 回升至 90.62%,如需严格单调可考虑合并这两箱

第 3 箱((0.82, 0.99])为极高风险区间:坏率高达 93.26%,策略上应重点管控

cyc_accountcount
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 3.5]3315.06%4728485.80%0.17
2(3.5, 10.5]315553.32%350280588.91%0.12
3(10.5, 11.5]41859.71%3338592.11%0.09
4(11.5, 16.5]153283.14%165136789.23%0.12
5(16.5, inf]1102100.00%80102292.74%0.08
合计6538675586389.68%0.12
2026-11-03T11:58:00.587844 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(85.80% → 92.74% → 92.74%),如需严格单调可考虑调整切点或合并分箱。IV=0.0379,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 92.11% 回落至 89.23%,如需严格单调可考虑合并这两箱

第 5 箱((16.5, inf])为极高风险区间:坏率高达 92.74%,策略上应重点管控

total_new_cnt_rate_3m
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0]137220.98%176119687.17%0.15
2(0, 0.02]56929.68%4552492.09%0.09
3(0.02, 0.11]319578.53%297289890.70%0.10
4(0.11, 0.17]70789.34%9361486.85%0.15
5(0.17, inf]697100.00%6463390.82%0.10
合计6540675586589.68%0.12
2026-11-03T11:58:00.717386 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.17% → 92.09% → 90.82%),如需严格单调可考虑调整切点或合并分箱。IV=0.0377,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 92.09% 回落至 90.70%,如需严格单调可考虑合并这两箱

第 2 箱((0, 0.02])为极高风险区间:坏率高达 92.09%,策略上应重点管控

mini_loan_rate_6m
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.2]248337.97%210227391.54%0.09
2(0.2, 0.37]224172.23%260198188.40%0.13
3(0.37, 0.48]73983.53%6867190.80%0.10
4(0.48, 0.57]54391.83%6248188.58%0.13
5(0.57, inf]534100.00%7545985.96%0.16
合计6540675586589.68%0.12
2026-11-03T11:58:00.841389 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(91.54% → 91.54% → 85.96%),如需严格单调可考虑调整切点或合并分箱。IV=0.0374,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 88.40% 回升至 90.80%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 0.2])为极高风险区间:坏率高达 91.54%,策略上应重点管控

apply_num_m12
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]154922.59%202134786.96%0.15
2(0.5, 1.5]311167.95%315279689.87%0.11
3(1.5, 2.5]103082.97%10592589.81%0.11
4(2.5, 3.5]43889.36%2940993.38%0.07
5(3.5, inf]730100.00%5967191.92%0.09
合计6858710614889.65%0.12
2026-11-03T11:58:00.973583 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(86.96% → 93.38% → 91.92%),如需严格单调可考虑调整切点或合并分箱。IV=0.0370,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 89.87% 回落至 89.81%,如需严格单调可考虑合并这两箱

第 4 箱((2.5, 3.5])为极高风险区间:坏率高达 93.38%,策略上应重点管控

loanearliestmonthsubcurmonth
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 47.5]70510.92%8262388.37%0.13
2(47.5, 53.5]36016.50%5630484.44%0.18
3(53.5, 85.5]201047.64%209180189.60%0.12
4(85.5, 93.5]66958.00%8458587.44%0.14
5(93.5, inf]2711100.00%235247691.33%0.09
合计6455666578989.68%0.12
2026-11-03T11:58:01.105752 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.37% → 91.33% → 91.33%),如需严格单调可考虑调整切点或合并分箱。IV=0.0369,该变量区分度偏弱但可用。

WOE 在第 4、5 箱之间穿越 0 线:前 4 箱 WOE>0(客户质量优于整体),后 1 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 88.37% 回落至 84.44%,如需严格单调可考虑合并这两箱

第 5 箱((93.5, inf])为极高风险区间:坏率高达 91.33%,策略上应重点管控

loan_rate_3m_12m
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.07]71611.13%10361385.61%0.17
2(0.07, 0.29]214544.46%208193790.30%0.11
3(0.29, 0.34]64854.53%7956987.81%0.14
4(0.34, 0.62]205286.42%177187591.37%0.09
5(0.62, inf]874100.00%9777788.90%0.12
合计6435664577189.68%0.12
2026-11-03T11:58:01.233219 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(85.61% → 91.37% → 88.90%),如需严格单调可考虑调整切点或合并分箱。IV=0.0364,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 90.30% 回落至 87.81%,如需严格单调可考虑合并这两箱

第 4 箱((0.34, 0.62])为极高风险区间:坏率高达 91.37%,策略上应重点管控

normal_repay_num
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 30.5]500873.02%531447789.40%0.12
2(30.5, 38.5]38178.58%2335893.96%0.06
3(38.5, 66.5]69188.66%7361889.44%0.12
4(66.5, 103.5]34693.70%2232493.64%0.07
5(103.5, inf]432100.00%6137185.88%0.16
合计6858710614889.65%0.12
2026-11-03T11:58:01.361556 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(89.40% → 93.96% → 85.88%),如需严格单调可考虑调整切点或合并分箱。IV=0.0363,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 89.40% 回升至 93.96%,如需严格单调可考虑合并这两箱

第 2 箱((30.5, 38.5])为极高风险区间:坏率高达 93.96%,策略上应重点管控

loanrecentlymonthsubcurmonth
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 1.5]273042.29%230250091.58%0.09
2(1.5, 3.5]233478.45%258207688.95%0.12
3(3.5, 4.5]53486.72%6946587.08%0.15
4(4.5, 6.5]52594.86%6446187.81%0.14
5(6.5, inf]332100.00%4528786.45%0.16
合计6455666578989.68%0.12
2026-11-03T11:58:01.485873 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(91.58% → 91.58% → 86.45%),如需严格单调可考虑调整切点或合并分箱。IV=0.0361,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 87.08% 回升至 87.81%,如需严格单调可考虑合并这两箱

第 1 箱((-inf, 1.5])为极高风险区间:坏率高达 91.58%,策略上应重点管控

setl_rem_prcp_paym_m3_num
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]476069.41%540422088.66%0.13
2(0.5, 1.5]131388.55%119119490.94%0.10
3(1.5, inf]785100.00%5173493.50%0.07
合计6858710614889.65%0.12
2026-11-03T11:58:01.600695 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性良好:随分栏增大,Bad Rate 单调上升(88.66% → 93.50%),WOE 单调下降(0.1026 → -0.5081),符合风控分箱预期。IV=0.0357,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 2 箱 WOE<0(风险高于平均)

第 3 箱((1.5, inf])为极高风险区间:坏率高达 93.50%,策略上应重点管控

nosvaing_997_bank_cnt_6m
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]118418.10%153103187.08%0.15
2(0.5, 3.5]346971.15%347312290.00%0.11
3(3.5, 4.5]56279.74%4651691.81%0.09
4(4.5, 6.5]61589.14%7653987.64%0.14
5(6.5, inf]710100.00%5365792.54%0.08
合计6540675586589.68%0.12
2026-11-03T11:58:01.716608 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.08% → 92.54% → 92.54%),如需严格单调可考虑调整切点或合并分箱。IV=0.0347,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 91.81% 回落至 87.64%,如需严格单调可考虑合并这两箱

第 5 箱((6.5, inf])为极高风险区间:坏率高达 92.54%,策略上应重点管控

query_success_rate_6m
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.19]65010.10%9155986.00%0.16
2(0.19, 0.41]149733.36%173132488.44%0.13
3(0.41, 0.69]219167.41%210198190.42%0.11
4(0.69, 0.77]42874.06%5037888.32%0.13
5(0.77, inf]1669100.00%140152991.61%0.09
合计6435664577189.68%0.12
2026-11-03T11:58:01.840142 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(86.00% → 91.61% → 91.61%),如需严格单调可考虑调整切点或合并分箱。IV=0.0338,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 90.42% 回落至 88.32%,如需严格单调可考虑合并这两箱

第 5 箱((0.77, inf])为极高风险区间:坏率高达 91.61%,策略上应重点管控

rh_ln_qy_fst_inte
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 675.5]271642.19%298241889.03%0.12
2(675.5, 688.5]41748.66%3138692.57%0.08
3(688.5, 706.5]94563.34%11583087.83%0.14
4(706.5, 724.5]151786.91%122139591.96%0.09
5(724.5, inf]843100.00%9874588.37%0.13
合计6438664577489.69%0.11
2026-11-03T11:58:02.034355 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(89.03% → 92.57% → 88.37%),如需严格单调可考虑调整切点或合并分箱。IV=0.0331,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 1→2 箱有局部反弹:Bad Rate 由 89.03% 回升至 92.57%,如需严格单调可考虑合并这两箱

第 2 箱((675.5, 688.5])为极高风险区间:坏率高达 92.57%,策略上应重点管控

nosvaing_998_bank_rate_6m
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, -3.75]95914.66%11584488.01%0.14
2(-3.75, -1.13]41621.02%2639093.75%0.07
3(-1.13, 0.26]232756.61%258206988.91%0.12
4(0.26, 0.76]179784.08%157164091.26%0.10
5(0.76, inf]1041100.00%11992288.57%0.13
合计6540675586589.68%0.12
2026-11-03T11:58:02.160275 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.01% → 93.75% → 88.57%),如需严格单调可考虑调整切点或合并分箱。IV=0.0330,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 93.75% 回落至 88.91%,如需严格单调可考虑合并这两箱

第 2 箱((-3.75, -1.13])为极高风险区间:坏率高达 93.75%,策略上应重点管控

total_997_cnt_1m
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]366255.99%428323488.31%0.13
2(0.5, 1.5]163280.95%139149391.48%0.09
3(1.5, 2.5]65790.99%6459390.26%0.11
4(2.5, inf]589100.00%4454592.53%0.08
合计6540675586589.68%0.12
2026-11-03T11:58:02.277088 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.31% → 92.53% → 92.53%),如需严格单调可考虑调整切点或合并分箱。IV=0.0321,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 91.48% 回落至 90.26%,如需严格单调可考虑合并这两箱

第 4 箱((2.5, inf])为极高风险区间:坏率高达 92.53%,策略上应重点管控

no_bank_rate_3_6
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.16]112817.25%14898086.88%0.15
2(0.16, 0.3]70328.00%6364091.04%0.10
3(0.3, 0.36]47535.26%6441186.53%0.16
4(0.36, 0.53]141856.94%143127589.92%0.11
5(0.53, inf]2816100.00%257255990.87%0.10
合计6540675586589.68%0.12
2026-11-03T11:58:02.389977 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(86.88% → 91.04% → 90.87%),如需严格单调可考虑调整切点或合并分箱。IV=0.0318,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 91.04% 回落至 86.53%,如需严格单调可考虑合并这两箱

第 2 箱((0.16, 0.3])为极高风险区间:坏率高达 91.04%,策略上应重点管控

nosvaing_bank_org_3m
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]205031.35%253179787.66%0.14
2(0.5, 1.5]176958.39%186158389.49%0.12
3(1.5, 2.5]113475.73%102103291.01%0.10
4(2.5, 4.5]109792.51%88100991.98%0.09
5(4.5, inf]490100.00%4644490.61%0.10
合计6540675586589.68%0.12
2026-11-03T11:58:02.508272 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(87.66% → 91.98% → 90.61%),如需严格单调可考虑调整切点或合并分箱。IV=0.0300,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 4→5 箱有局部反弹:Bad Rate 由 91.98% 回落至 90.61%,如需严格单调可考虑合并这两箱

第 4 箱((2.5, 4.5])为极高风险区间:坏率高达 91.98%,策略上应重点管控

c_bank_credit_used_limit
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.09]319448.84%373282188.32%0.13
2(0.09, 0.73]212981.39%187194291.22%0.10
3(0.73, 0.83]34286.62%2531792.69%0.08
4(0.83, 0.99]47693.90%5542188.45%0.13
5(0.99, inf]399100.00%3536491.23%0.10
合计6540675586589.68%0.12
2026-11-03T11:58:02.626279 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(88.32% → 92.69% → 91.23%),如需严格单调可考虑调整切点或合并分箱。IV=0.0291,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 92.69% 回落至 88.45%,如需严格单调可考虑合并这两箱

第 3 箱((0.73, 0.83])为极高风险区间:坏率高达 92.69%,策略上应重点管控

c_bank_rate_1_6
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.12]432066.06%472384889.07%0.12
2(0.12, 0.23]59975.21%5254791.32%0.10
3(0.23, 0.33]37880.99%2135794.44%0.06
4(0.33, 0.59]82493.59%9273288.83%0.13
5(0.59, inf]419100.00%3838190.93%0.10
合计6540675586589.68%0.12
2026-11-03T11:58:02.751310 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(89.07% → 94.44% → 90.93%),如需严格单调可考虑调整切点或合并分箱。IV=0.0281,该变量区分度偏弱但可用。

WOE 在第 1、2 箱之间穿越 0 线:前 1 箱 WOE>0(客户质量优于整体),后 4 箱 WOE<0(风险高于平均)

第 3→4 箱有局部反弹:Bad Rate 由 94.44% 回落至 88.83%,如需严格单调可考虑合并这两箱

第 3 箱((0.23, 0.33])为极高风险区间:坏率高达 94.44%,策略上应重点管控

no_bank_cyc_used_limit
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.24]408062.39%427365389.53%0.12
2(0.24, 0.27]33067.43%4728385.76%0.17
3(0.27, 0.33]58876.42%4953991.67%0.09
4(0.33, 0.4]49083.91%3345793.27%0.07
5(0.4, inf]1052100.00%11993388.69%0.13
合计6540675586589.68%0.12
2026-11-03T11:58:02.879177 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

关键结论

单调性欠佳:Bad Rate 随分栏存在波动(89.53% → 93.27% → 88.69%),如需严格单调可考虑调整切点或合并分箱。IV=0.0278,该变量区分度偏弱但可用。

WOE 在第 2、3 箱之间穿越 0 线:前 2 箱 WOE>0(客户质量优于整体),后 3 箱 WOE<0(风险高于平均)

第 2→3 箱有局部反弹:Bad Rate 由 85.76% 回升至 91.67%,如需严格单调可考虑合并这两箱

第 4 箱((0.33, 0.4])为极高风险区间:坏率高达 93.27%,策略上应重点管控

13 压力测试

注:验证压力情境下模型的有效性和区分能力稳定性:a.坏浓度更高样本 b.时间外样本 c.特定特征客群 d.极端时间样本。当前流程自动生成情景1(贡献度最高的变量调用失败,20261026 批注2)与情景2(客群资质变差——跨时间样本剔除一半好样本,批注16)。

测试情景1:贡献度最高的变量调用失败

变量贡献度

变量贡献度
odgt0_m12_num23.0000
m_c_over_due_cnt7.0000
odgt0_m12_mob_pc1.0000
last6average0.0000
loanearliestmonthsubcurmonth0.0000
cur_od_loan_amt0.0000
rh_cr_bal_pro0.0000
no_bank_rate_3_60.0000
rh_ln_qy_fst_inte0.0000
ps_od_norm_int_amt0.0000
失效变量贡献度口径值KS(前 → 后)AUC(前 → 后)KS 衰减AUC 衰减
odgt0_m12_num293.92910.2313 → 0.17880.6238 → 0.6026-0.0525-0.0212

注:失效处理口径:LR 将该变量 WOE 置 0(中性箱)后按系数重算概率;树模型将该特征置缺失重打分;评估样本 = 跨时间验证样本。衰减幅度反映模型对头部变量的依赖程度。

区分度验证 — LightGBM @ 跨时间样本

KS = 0.2313,GINI = 0.2476

2026-11-03T11:58:03.010557 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-11-03T11:58:03.118427 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

测试情景2:客群资质变差

指标数值
样本量1065 → 995(剔除好样本 70)
KS(前 → 后)0.2313 → 0.2099(Δ -0.0214)
AUC(前 → 后)0.6238 → 0.6033(Δ -0.0205)
GINI(后)0.2066
2026-11-03T11:58:03.229748 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-11-03T11:58:03.413896 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

注:情景2 模拟客群资质变差的压力情形(好客户变少、坏客户占比升高):随机剔除一半好样本(固定 seed=42,可复现)后复算 KS/AUC。