风控模型开发验证报告

实验 ID: 20261027-094646-dd41 · 生成时间: 2026-10-27 09:47 (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_20261027-094646-dd41
开发完成日期2026-10-27
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preloan_approval
模型应用频率(人工填写)
模型开发背景及业务用途描述Local MVP full pipeline run.
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LR + WOE 评分卡
目标变量gbflag
实验 ID20261027-094646-dd41
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2021-10-17 ~ 2026-04-30
预留验证样本窗口2024-04-30 ~ 2026-04-30
跨时间验证样本窗口2021-12-31 ~ 2026-06-30
PSI时点样本窗口2023-11-18 ~ 2026-07-31

模型好坏定义

项目内容
目标变量gbflag
坏样本(BAD=1)占比8.03%
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
2create_time主键/时间/口径字段,不入模
3gbflag 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

开发时点 — 样本排除

类别样本数
全部样本115682
排除:目标变量为空0
排除后样本115682

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11568210638992938.03%100.00%
开发样本380883493731518.27%32.92%
测试样本253912329121008.27%21.95%
训练+测试样本634795822852518.27%54.87%
跨时间验证样本347303229824327.00%30.02%
PSI时点样本17473158631610—15.10%

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582283493760.00%2329140.00%
坏5251315160.01%210039.99%
合计634793808860.00%2539140.00%

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
due_day_my11.61160.00%46.32%0.0972
last_od_to_cur_freq0.19280.02%70.20%0.4978
max_od_days0.18700.02%25.81%0.5557
last_loan_setl_tnr0.16690.02%18.96%0.3034
end_loan_to_cur_freq0.14950.02%6.32%0.6425
apprv_amt_day0.14300.02%79.67%0.6396
cre_l60_used_limit_rate_min0.12720.06%10.21%0.7274
ps_perd_no0.12440.00%13.10%0.6425
loan_os_prcp_credit_amount_m30.12220.07%9.66%0.5204
cre_l90_used_limit_amt_min0.11600.06%14.62%0.7390
shd_pre_repay_due_rate0.11041.11%16.57%0.5562
odgt0_m12_subcurmonth0.10940.07%34.86%0.7415
cre_now_l60_used_amt_m_min0.10490.06%74.02%0.6396
cre_now_l60_limit_left_d_max0.10040.48%67.92%0.7660
loan_os_prcp_credit_amount_m60.09440.07%2.98%0.5204
f_ps_od_days_max0.09320.02%53.63%0.5202
suc_apprv_num_m10.09090.02%79.23%0.6800
acc_ps_perd_no0.08750.02%25.81%0.4281
last_od_num_m60.08590.02%78.72%0.6476
od_amt_m6_loan_os_prcp_m6_pc0.07980.07%93.14%0.6644
loan_os_prcp_credit_amount_m120.07950.07%0.23%0.2371
ps_int_amt_mon0.07750.02%5.50%0.7503
cre_l30_l90_used_amt_d_max0.07720.34%21.39%0.6593
available_limit_amt0.07710.02%32.48%0.8089
setl_prcp_ratio0.07680.02%21.83%0.6593
orig_prcp_avg_m30.07630.02%60.04%0.5989
cre_l30_l90_used_amt_m_max0.07310.06%21.48%0.6810
cre_now_l30_used_amt_d_min0.07105.71%86.91%0.9733
shd_max_dpd0.06901.05%74.62%0.3801
cre_l60_used_limit_rate_max0.06900.06%6.42%0.7274
od_amt_m12_loan_os_prcp_m12_pc0.06490.07%92.39%0.6644
repay_amount_m3_loan_os_prcp_pc0.06340.07%38.42%0.6366
t_ps_od_days_max0.06070.02%54.18%0.5455
apprv_num_m10.05950.02%73.93%0.7533
s_ps_od_days_max0.05940.02%54.12%0.5455
shd_l30_max_dpd0.05762.96%71.65%0.7038
apprv_num_m30.05720.02%52.89%0.7533
useratemore50maxcontimonsm120.05300.07%10.71%0.6543
cre_l30_limit_left_amt_min0.05040.06%4.18%0.7875
ps_int_amt_sum0.04650.02%1.21%0.7396
cre_now_l60_limit_left_d_min0.04565.74%11.24%0.8178
avg_tnr0.04560.02%30.34%0.6841
odgt0_m12_mob_pc0.04470.07%33.87%0.6158
cre_now_l60_used_amt_d_min0.042910.25%81.68%0.7521
max_loan_os_prcp_m120.04110.07%4.04%0.7503
odgt0_m12_num0.04070.07%34.86%0.7141
useratemore70_m6_num0.04050.07%28.42%0.6941
cre_now_l60_used_amt_m_max0.04000.06%11.21%0.6810
suc_apprv_num_m60.03950.02%34.85%0.7214
cre_now_p1m_used_rate_d0.03915.64%4.54%0.9733
useratemore70monthsubcurmonth0.03840.07%35.85%0.6077
orig_prcp_avg_m120.03720.02%3.24%0.6695
cre_l30_l90_used_amt_m_min0.03690.06%78.40%0.5989
shd_pre_repay_amt_rate0.03380.08%1.29%0.6789
cre_usedrate_u30_daydelta_max0.033012.49%83.72%0.6366
repay_amount_m6_all_shd_amt_pc0.03210.07%3.66%0.7885
ps_int_amt_avg0.03100.02%1.21%0.6095
cre_now_l30_used_amt_d_max0.02990.52%14.27%0.6000
cre_now_l30_limit_left_d_min0.02894.23%14.95%0.3058
repay_amount_m6_loan_os_prcp_pc0.02820.07%54.31%0.5912
shd_l60_min_dpd0.02782.50%51.07%0.7466
min_apply_amt0.02730.02%10.57%0.4938
useratemore90maxcontimonsm120.02690.07%24.43%0.6895
cre_now_l90_limit_left_d_min0.02556.91%9.77%0.8178
repay_amount_all_shd_amt_pc0.02540.07%5.67%0.6536
shd_instm_amt_sum0.02480.08%0.03%0.7396
useratemore90_m3_num0.02400.07%61.13%0.6941
cre_l30_l90_usedrate_m_max0.02280.06%26.85%0.7107
odgt0_m3_num0.02270.07%53.85%0.7141
cre_now_l60_usedrate_m_max0.02260.06%13.38%0.7107
m_s_ps_instm_amt0.02160.02%70.25%0.4301
min_tnr0.02150.02%53.18%0.6841
cre_now_p3m_used_rate_d0.021211.82%1.60%0.7521
setl_rem_prcp_paym_last_month_lo0.02030.07%87.85%0.6536
shd_l180_instm_amt_sum0.02000.08%0.11%0.7458
last_clear_num_m30.02000.02%57.66%0.7214
loan_os_prcp_gt0_m12_num_pc0.01960.07%61.32%0.5912
ps_prcp_amt_mon0.01830.02%4.60%0.7458
cre_now_p1m_used_rate_m0.01740.93%4.84%0.7553
cre_now_p2m_used_rate_m0.01581.66%2.23%0.7553
shd_l30_pre_repay_due_rate0.01533.11%48.56%0.7855
end_loan_to_rep_loan_freq0.01330.02%6.97%0.4280
shd_pre_repay_cnt0.01330.08%1.22%0.5711
shd_remain_actual_cnt0.01250.08%7.33%0.6008
account_age0.01240.07%4.74%0.7206
shd_l90_ps_due_cnt0.01180.08%17.68%0.6782
cre_now_l30_limit_left_d_max0.01160.72%79.30%0.7660
min_apprv_amt0.01100.02%29.17%0.5699
last_clear_num_m10.01000.02%79.20%0.7141
shd_min_dpd0.00941.06%26.93%0.7113
sum_apply_amt0.00870.02%10.45%0.4938
account_age_loan_day0.00840.07%2.11%0.7206
shd_remain_actual_rate0.00820.41%66.46%0.7855
max_int_rat0.00800.02%34.15%0.4054
last_apply_freq0.00800.02%87.14%0.6518
max_od_due0.00690.07%74.02%0.7415
setl_prcp_mon_ratio0.00610.07%6.05%0.4346
grade_limit_amount0.00510.02%38.22%0.4065
shd_l180_max_dpd0.00431.69%57.75%0.7885
cre_usedrate_u30_daydelta_min0.004012.49%81.81%0.4730

注:共 125 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)

due_day_my — IV 19.9031,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 30.5]349373493700.00%8.7486
2(30.5, inf]315103151100.00%-11.1544
last_od_to_cur_freq — IV 0.1995,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]4011321679519.82%-1.0084
2(1.5, 3.5]260823722369.05%-0.0983
3(3.5, inf]314612934221196.74%0.2221
f_ps_od_days_max — IV 0.1798,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]2042218144227811.15%-0.3309
2(0.5, inf]17658167868724.94%0.5516
ovd_days_d3_num_m1 — IV 0.1789,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]346763222824487.06%0.1716
2(0.5, inf]3404270270220.62%-1.0581
suc_apprv_num_m1 — IV 0.1594,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]301692728728829.55%-0.1580
2(0.5, 1.5]478745911964.09%0.7478
3(1.5, inf]31243052722.30%1.3409
last_loan_setl_tnr — IV 0.1487,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]723969942453.38%0.9456
2(0.5, 3.5]11453105429117.95%0.0426
3(3.5, inf]1938817394199410.28%-0.2400
cre_l60_used_limit_rate_min — IV 0.1478,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.019700000062584877]444143061353.04%1.0570
2(0.019700000062584877, 0.09174999967217445]281326561575.58%0.4228
3(0.09174999967217445, 0.2787500023841858]732368075167.05%0.1741
4(0.2787500023841858, 0.9161500036716461]188991715717429.22%-0.1181
5(0.9161500036716461, inf]4589398960013.07%-0.5111
apprv_amt_day — IV 0.1475,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 11.0]303442745228929.53%-0.1555
2(11.0, inf]773674782583.34%0.9608
ps_perd_no — IV 0.1442,4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]499148451462.93%1.0963
2(1.5, 2.5]317029961745.49%0.4402
3(2.5, 3.5]322530072186.76%0.2184
4(3.5, inf]267022408926139.79%-0.1846
cre_l90_used_limit_amt_min — IV 0.1336,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 21.640000343322754]561854082103.74%0.8430
2(21.640000343322754, 748.7099914550781]807675265506.81%0.2107
3(748.7099914550781, 8516.22998046875]201721835418189.01%-0.0934
4(8516.22998046875, 14597.384765625]2267199027712.22%-0.4336
5(14597.384765625, inf]1932163729515.27%-0.6919
cre_now_l60_used_amt_m_min — IV 0.1300,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.7799999713897705]282012545827439.73%-0.1775
2(1.7799999713897705, inf]986494574074.13%0.7402
apprv_num_m1 — IV 0.1224,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]281522542527279.69%-0.1734
2(0.5, 1.5]535051072434.54%0.6394
3(1.5, inf]457843981803.93%0.7900
orig_prcp_avg_m3 — IV 0.1156,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 23.449999809265137]2286420513235110.28%-0.2397
2(23.449999809265137, inf]15216144177995.25%0.4869
s_ps_od_days_max — IV 0.1150,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]2061018430218010.58%-0.2713
2(0.5, inf]17470165009705.55%0.4279
last_od_num_m6 — IV 0.1092,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]299772791920586.87%0.2016
2(0.5, inf]81037011109213.48%-0.5465
available_limit_amt — IV 0.1036,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.03499999921768904]1238510918146711.85%-0.3987
2(0.03499999921768904, 27316.265625]237712218315886.68%0.2309
3(27316.265625, inf]19241829954.94%0.5517
cre_now_l30_used_amt_d_min — IV 0.1011,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.0000500082969666]312112827329389.41%-0.0980
2(1.0000500082969666, 1.382200002670288]19331860733.78%0.8757
3(1.382200002670288, inf]27682687812.93%1.1395
acc_ps_perd_no — IV 0.0987,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]98278680114711.67%-0.3821
2(0.5, 7.5]132211211111108.40%-0.0162
3(7.5, 23.5]846178965656.68%0.2313
4(23.5, 65.5]462543732525.45%0.4478
5(65.5, inf]19461870763.91%0.7970
apprv_num_m3 — IV 0.0951,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]2013918040209910.42%-0.2548
2(0.5, 2.5]1020095756256.13%0.3232
3(2.5, inf]774173154265.50%0.4373
ps_int_amt_mon — IV 0.0946,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.475000023841858]23542275793.36%0.9543
2(1.475000023841858, 19.704999923706055]933887056336.78%0.2152
3(19.704999923706055, 132.375]194131782715868.17%0.0136
4(132.375, 296.6699981689453]4803426753611.16%-0.3314
5(296.6699981689453, inf]2172185631614.55%-0.6355
max_od_days — IV 0.0920,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 5.5]313952916222337.11%0.1636
2(5.5, inf]6685576891713.72%-0.5670
setl_prcp_ratio — IV 0.0909,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.035099999979138374]832579603654.38%0.6763
2(0.035099999979138374, inf]297552697027859.36%-0.1355
end_loan_to_cur_freq — IV 0.0902,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -22.5]299592716227979.34%-0.1327
2(-22.5, inf]812177683534.35%0.6854
loan_os_prcp_credit_amount_m12 — IV 0.0894,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 16.18000030517578]316430581063.35%0.9564
2(16.18000030517578, 44.21000099182129]603356713626.00%0.3458
3(44.21000099182129, 77.875]144531322312308.51%-0.0308
4(77.875, 95.76499938964844]663560086279.45%-0.1458
5(95.76499938964844, inf]7775695182410.60%-0.2732
cre_l30_l90_used_amt_m_max — IV 0.0868,4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 35.010000228881836]863382194144.80%0.5828
2(35.010000228881836, 350.3800048828125]631158194927.80%0.0649
3(350.3800048828125, 1951.3800048828125]153141391613989.13%-0.1075
4(1951.3800048828125, inf]7807696184610.84%-0.2980
avg_od_amt_m6_orig_prcp_pc — IV 0.0815,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5699999928474426]355683286627027.60%0.0927
2(0.5699999928474426, inf]2492204544717.94%-0.8851
od_amt_m6_loan_os_prcp_m6_pc — IV 0.0804,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.38500000536441803]355693286527047.60%0.0920
2(0.38500000536441803, inf]2491204644517.86%-0.8801
cre_now_l60_used_amt_d_min — IV 0.0761,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.000950038433075]279432520527389.80%-0.1106
2(1.000950038433075, 1.4549500346183777]235622351215.14%0.5858
3(1.4549500346183777, inf]388637151714.40%0.7481
cre_now_l30_limit_left_d_max — IV 0.0708,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.9999499917030334]721468873274.53%0.6418
2(0.9999499917030334, inf]306002779828029.16%-0.1110
cre_l30_l90_used_amt_d_max — IV 0.0705,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.9735499918460846]291702645527159.31%-0.1269
2(0.9735499918460846, inf]879083584324.91%0.5590
shd_pre_repay_due_rate — IV 0.0693,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.868800014257431]1047399205535.28%0.4846
2(0.868800014257431, inf]271912461825739.46%-0.1439
shd_max_dpd — IV 0.0690,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]95658427113811.90%-0.4008
2(0.5, inf]281222613419887.07%0.1731
loan_os_prcp_gt0_m3_num_pc — IV 0.0681,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 68.16500091552734]335432521023.04%1.0563
2(68.16500091552734, inf]347063165930478.78%-0.0649
loan_os_prcp_credit_amount_m6 — IV 0.0668,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 33.795000076293945]812777343934.84%0.5739
2(33.795000076293945, inf]299332717727569.21%-0.1171
od_amt_m12_loan_os_prcp_m12_pc — IV 0.0658,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.0950000025331974]352223253226907.64%0.0870
2(0.0950000025331974, inf]2838237945916.17%-0.7603
max_od_due — IV 0.0638,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]98078661114611.69%-0.3832
2(0.5, inf]282532625020037.09%0.1673
odgt0_m12_subcurmonth — IV 0.0635,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]98048659114511.68%-0.3825
2(0.5, inf]282562625220047.09%0.1669
cre_l60_used_limit_rate_max — IV 0.0614,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.22405000030994415]426540422235.23%0.4918
2(0.22405000030994415, 1.2138000130653381]265202445720637.78%0.0672
3(1.2138000130653381, inf]7280641686411.87%-0.4005
shd_l30_max_dpd — IV 0.0576,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]278582582320357.30%0.1535
2(0.5, inf]91018026107511.81%-0.3769
ps_int_amt_sum — IV 0.0558,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 209.51499938964844]317330021715.39%0.4594
2(209.51499938964844, 815.2900085449219]1035696177397.14%0.1600
3(815.2900085449219, 1891.7449951171875]1042996138167.82%0.0605
4(1891.7449951171875, 10533.93994140625]113161027610409.19%-0.1154
5(10533.93994140625, inf]2806242238413.69%-0.5642
useratemore50maxcontimonsm12 — IV 0.0545,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]407638891874.59%0.6291
2(0.5, 2.5]428940352545.92%0.3597
3(2.5, inf]296952698727089.12%-0.1066
avg_tnr — IV 0.0524,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 3.5]24362340963.94%0.7876
2(3.5, 5.5]367234422306.26%0.2998
3(5.5, 6.5]266024721887.07%0.1704
4(6.5, 10.5]126711164510268.10%0.0233
5(10.5, inf]166411503116109.67%-0.1721
cre_now_l60_used_amt_m_max — IV 0.0511,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 83.30500030517578]514849042444.74%0.5951
2(83.30500030517578, 2003.010009765625]240882208320058.32%-0.0063
3(2003.010009765625, inf]8829792890110.21%-0.2309
shd_l30_pre_repay_due_rate — IV 0.0506,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.8838499784469604]6969614482511.84%-0.3781
2(0.8838499784469604, 4.9175498485565186]279212573621857.83%0.0803
3(4.9175498485565186, inf]20131914994.92%0.5758
cre_usedrate_u30_daydelta_max — IV 0.0468,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 57.5]282927001294.56%0.6900
2(57.5, 85.5]211219921205.68%0.4582
3(85.5, inf]283912574126509.33%-0.0776
odgt0_m12_num — IV 0.0464,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]1326811926134210.11%-0.2212
2(0.5, 1.5]773771006378.23%0.0054
3(1.5, 2.5]553450984367.88%0.0532
4(2.5, 5.5]832177475746.90%0.1967
5(5.5, inf]320030401605.00%0.5387
cre_l30_limit_left_amt_min — IV 0.0459,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -132.05500030517578]7225636985611.85%-0.3986
2(-132.05500030517578, inf]308402854622947.44%0.1157
max_loan_os_prcp_m12 — IV 0.0455,4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1747.2849731445312]410138602415.88%0.3679
2(1747.2849731445312, 13572.5849609375]252862332819587.74%0.0720
3(13572.5849609375, 20848.58984375]361932783419.42%-0.1426
4(20848.58984375, inf]5054444560912.05%-0.4180
odgt0_m12_mob_pc — IV 0.0442,4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4.164999961853027]1327211930134210.11%-0.2208
2(4.164999961853027, 20.130000114440918]125321151210208.14%0.0179
3(20.130000114440918, 57.670000076293945]1000393296746.74%0.2219
4(57.670000076293945, inf]225321401135.02%0.5355
setl_rem_prcp_paym_m3_num — IV 0.0433,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]320942924028548.89%-0.0789
2(0.5, inf]596656712954.94%0.5504
orig_prcp_avg_m12 — IV 0.0431,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1031.9722290039062]11498107357636.64%0.2381
2(1031.9722290039062, 1871.875]692564055207.51%0.1051
3(1871.875, 3819.857177734375]872779937348.41%-0.0181
4(3819.857177734375, 13787.5]890980418689.74%-0.1798
5(13787.5, inf]2021175626513.11%-0.5149
useratemore70_m6_num — IV 0.0429,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]10817101466716.20%0.3103
2(0.5, 1.5]443541153207.22%0.1484
3(1.5, 4.5]144121310413089.08%-0.1013
4(4.5, 5.5]346831323369.69%-0.1734
5(5.5, inf]4928441451410.43%-0.2554
useratemore70monthsubcurmonth — IV 0.0416,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4.5]254292310223279.15%-0.1104
2(4.5, 9.5]603755644737.83%0.0593
3(9.5, inf]659462453495.29%0.4787
suc_apprv_num_m6 — IV 0.0404,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]220111992520869.48%-0.1492
2(1.5, 3.5]781272405727.32%0.1323
3(3.5, inf]825777654925.96%0.3530
m_s_ps_instm_amt — IV 0.0363,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 33.55999946594238]268902491619747.34%0.1295
2(33.55999946594238, 1121.7000122070312]9235829893710.15%-0.2249
3(1121.7000122070312, inf]1955171623912.23%-0.4347
shd_pre_repay_amt_rate — IV 0.0360,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.6238499879837036]12637118148236.51%0.2584
2(0.6238499879837036, 0.7506499886512756]863179446877.96%0.0421
3(0.7506499886512756, inf]167911515216399.76%-0.1816
useratemore90maxcontimonsm12 — IV 0.0350,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]929887235756.18%0.3136
2(0.5, 1.5]690363495548.03%0.0332
3(1.5, 4.5]147501346112898.74%-0.0598
4(4.5, 7.5]465241884649.97%-0.2056
5(7.5, inf]2457219026710.87%-0.3013
ps_int_amt_avg — IV 0.0344,4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4.333299875259399]408338572265.54%0.4312
2(4.333299875259399, 13.080849647521973]12227113149137.47%0.1111
3(13.080849647521973, 38.24515151977539]143511310612458.68%-0.0520
4(38.24515151977539, inf]7419665376610.32%-0.2443
repay_amount_m6_all_shd_amt_pc — IV 0.0344,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 93.98500061035156]132571194613119.89%-0.1961
2(93.98500061035156, 99.48500061035156]197431819315507.85%0.0571
3(99.48500061035156, inf]506047722885.69%0.4018
cre_l30_l90_used_amt_m_min — IV 0.0320,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.6299999952316284]298562718526718.95%-0.0853
2(0.6299999952316284, inf]820977304795.84%0.3756
cre_now_p1m_used_rate_m — IV 0.0308,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -4.999999873689376e-05]728868704185.74%0.3962
2(-4.999999873689376e-05, inf]304462773527118.90%-0.0779
last_clear_num_m3 — IV 0.0305,5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]219561990820489.33%-0.1317
2(0.5, 1.5]704365255187.35%0.1275
3(1.5, 2.5]341731772407.02%0.1771
4(2.5, 3.5]203219031296.35%0.2854
5(3.5, inf]363234172155.92%0.3599
useratemore90_m3_num — IV 0.0302,4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]232662150117657.59%0.0942
2(0.5, 1.5]611756464717.70%0.0781
3(1.5, 2.5]397936233568.95%-0.0856
4(2.5, inf]4698414155711.86%-0.3996
min_apply_amt — IV 0.0286,4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1135.0]3388301037811.16%-0.3311
2(1135.0, 4180.0]179831641315708.73%-0.0589
3(4180.0, 7300.0]922184907317.93%0.0463
4(7300.0, inf]748870174716.29%0.2953
repay_amount_m3_loan_os_prcp_pc — IV 0.0283,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -4999.0]480445542505.20%0.4966
2(-4999.0, inf]332563035728998.72%-0.0571
shd_l90_ps_due_cnt — IV 0.0275,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 2.5]248623831034.14%0.7357
2(2.5, inf]355733252730468.56%-0.0374
ps_prcp_amt_mon — IV 0.0267,4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 38.48500061035156]206919661034.98%0.5431
2(38.48500061035156, 955.4700012207031]225732080117727.85%0.0569
3(955.4700012207031, 4656.4951171875]113801034210389.12%-0.1070
4(4656.4951171875, inf]2058182123711.52%-0.3669
cre_now_l30_limit_left_d_min — IV 0.0262,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.997050017118454]6136544369311.29%-0.3306
2(0.997050017118454, inf]303392797523647.79%0.0793
shd_instm_amt_sum — IV 0.0258,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 17727.4052734375]154541433311217.25%0.1427
2(17727.4052734375, 134568.0234375]178261634114858.33%-0.0074
3(134568.0234375, inf]4779423654311.36%-0.3514
loan_os_prcp_gt0_m12_num_pc — IV 0.0242,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 57.73500061035156]329831381604.85%0.5705
2(57.73500061035156, 72.08000183105469]263124172148.13%0.0186
3(72.08000183105469, inf]321312935627758.64%-0.0469
shd_l180_instm_amt_sum — IV 0.0229,4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 5366.869873046875]147451367710687.24%0.1442
2(5366.869873046875, 34552.490234375]187101713715738.41%-0.0174
3(34552.490234375, 55994.666015625]2458220725110.21%-0.2318
4(55994.666015625, inf]2146188925711.98%-0.4110
odgt0_m3_num — IV 0.0227,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]300542771923357.77%0.0684
2(1.5, 2.5]536648824849.02%-0.0945
3(2.5, inf]2640231033012.50%-0.4598
cre_l30_l90_usedrate_m_max — IV 0.0224,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.025150000117719173]11879110807996.73%0.2240
2(0.025150000117719173, 0.0799499973654747]331030532577.76%0.0693
3(0.0799499973654747, inf]228762078220949.15%-0.1105
cre_now_l60_limit_left_d_min — IV 0.0217,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.9717499911785126]8039718685310.61%-0.2558
2(0.9717499911785126, inf]278622569321697.78%0.0850
min_tnr — IV 0.0215,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 5.0]229492126016897.36%0.1267
2(5.0, inf]151311367014619.66%-0.1699
cre_now_l60_usedrate_m_max — IV 0.0213,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.039750000461936]772672484786.19%0.3134
2(0.039750000461936, inf]303392766726728.81%-0.0681
tothismonthodcontireducenumm12 — IV 0.0209,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]328993004428558.68%-0.0521
2(0.5, inf]516148672945.70%0.4009
cre_now_l90_limit_left_d_min — IV 0.0204,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.9937500059604645]9689869799210.24%-0.2183
2(0.9937500059604645, inf]257672378119867.71%0.0935
last_clear_num_m1 — IV 0.0203,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]301612750926528.79%-0.0667
2(0.5, 1.5]491645943226.55%0.2520
3(1.5, inf]300328271765.86%0.3705
setl_rem_prcp_paym_last_month_lo — IV 0.0202,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -4999.0]289927571424.90%0.5604
2(-4999.0, inf]351613215430078.55%-0.0361
cre_now_p2m_used_rate_m — IV 0.0190,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -0.00014999999621068127]1033696486886.66%0.2388
2(-0.00014999999621068127, inf]271212469924228.93%-0.0797
tothismonthodcontiaddnumm12 — IV 0.0187,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]330573046325947.85%0.0576
2(0.5, inf]5003444855511.09%-0.3245
shd_pre_repay_cnt — IV 0.0181,4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 25.5]956088766847.15%0.1575
2(25.5, 237.5]190571752515328.04%0.0314
3(237.5, 362.5]253723142238.79%-0.0661
4(362.5, inf]6905619571010.28%-0.2395
last_apply_freq — IV 0.0169,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 6.5]333023067926237.88%0.0533
2(6.5, inf]4778425152711.03%-0.3182
t_ps_od_days_max — IV 0.0168,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 5.5]359753309328828.01%0.0349
2(5.5, inf]2105183726812.73%-0.4810
cre_now_p3m_used_rate_d — IV 0.0161,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.40674999356269836]436340892746.28%0.3566
2(0.40674999356269836, inf]292222656226609.10%-0.0451
shd_perd_max — IV 0.0161,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 8.5]20471951964.69%0.6061
2(8.5, inf]360123295930538.48%-0.0265
cre_now_l30_used_amt_d_max — IV 0.0151,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.6786499917507172]737668984786.48%0.2678
2(0.6786499917507172, inf]305152784626698.75%-0.0566
cre_now_p1m_used_rate_d — IV 0.0140,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.0000500082969666]302512754327088.95%-0.0478
2(1.0000500082969666, inf]568953173726.54%0.2924
account_age — IV 0.0131,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 29.5]253362339819387.65%0.0853
2(29.5, inf]127241151312119.52%-0.1537
shd_remain_actual_cnt — IV 0.0129,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 3.5]712966624676.55%0.2522
2(3.5, inf]309302824826828.67%-0.0512
shd_pre_repay_cnt_rate — IV 0.0124,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]523249043286.27%0.2991
2(0.5, inf]328273000628218.59%-0.0414
setl_prcp_mon_ratio — IV 0.0121,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.0011139735579490662]2888259229610.25%-0.2359
2(0.0011139735579490662, 0.09751652926206589]328223010627168.27%-0.0002
3(0.09751652926206589, inf]235022131375.83%0.3764
cre_usedrate_u30_daydelta_min — IV 0.0116,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 2.5]273532486924849.08%-0.0474
2(2.5, inf]597955644156.94%0.2446
grade_limit_amount — IV 0.0103,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 84500.0]2178194523310.70%-0.2840
2(84500.0, 180500.0]175811605715248.67%-0.0511
3(180500.0, inf]183211692813937.60%0.0916
shd_remain_actual_rate — IV 0.0095,3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.41544999182224274]378835512376.26%0.3042
2(0.41544999182224274, 0.7352499961853027]485144613908.04%0.0342
3(0.7352499961853027, inf]292942677425208.60%-0.0396
suc_apply_trench_num — IV 0.0091,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]355033264028638.06%0.0277
2(1.5, inf]2577229028711.14%-0.3291
cre_now_l60_limit_left_d_max — IV 0.0088,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -0.013299999758601189]2253199925411.27%-0.3455
2(-0.013299999758601189, inf]356513277728748.06%0.0255
max_int_rat — IV 0.0079,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.2142000049352646]372734912366.33%0.2882
2(0.2142000049352646, inf]343533143929148.48%-0.0274
repay_amount_m6_loan_os_prcp_pc — IV 0.0070,2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -4999.0]682863534756.96%0.1877
2(-4999.0, inf]312322855826748.56%-0.0374

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 125 个;完整数据见 feature_report.json。

04 单变量分析(粗分箱)

due_day_my — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 30.5]349373493700.00%8.7486
2(30.5, inf]315103151100.00%-11.1544
2026-10-27T09:47:25.410188 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(0.00% → 100.00%),WOE 单调下降(8.7486 → -11.1544),符合风控分箱预期。IV=19.9031,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((30.5, inf])为极高风险区间:坏率高达 100.00%,策略上应重点管控

f_ps_od_days_max — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]2042218144227811.15%-0.3309
2(0.5, inf]17658167868724.94%0.5516
2026-10-27T09:47:25.514177 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(11.15% → 4.94%),WOE 单调上升(-0.3309 → 0.5516),符合风控分箱预期。IV=0.1798,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 1 箱((-inf, 0.5])为坏率最高区间:坏率高达 11.15%,可作为拒绝或重点风控策略的阈值参考

suc_apprv_num_m1 — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]301692728728829.55%-0.1580
2(0.5, 1.5]478745911964.09%0.7478
3(1.5, inf]31243052722.30%1.3409
2026-10-27T09:47:25.627278 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(9.55% → 2.30%),WOE 单调上升(-0.1580 → 1.3409),符合风控分箱预期。IV=0.1594,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])为坏率最高区间:坏率高达 9.55%,可作为拒绝或重点风控策略的阈值参考

cre_l60_used_limit_rate_min — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.019700000062584877]444143061353.04%1.0570
2(0.019700000062584877, 0.09174999967217445]281326561575.58%0.4228
3(0.09174999967217445, 0.2787500023841858]732368075167.05%0.1741
4(0.2787500023841858, 0.9161500036716461]188991715717429.22%-0.1181
5(0.9161500036716461, inf]4589398960013.07%-0.5111
2026-10-27T09:47:25.751497 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(3.04% → 13.07%),WOE 单调下降(1.0570 → -0.5111),符合风控分箱预期。IV=0.1478,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 5 箱((0.9161500036716461, inf])为坏率最高区间:坏率高达 13.07%,可作为拒绝或重点风控策略的阈值参考

apprv_amt_day — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 11.0]303442745228929.53%-0.1555
2(11.0, inf]773674782583.34%0.9608
2026-10-27T09:47:25.869039 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(9.53% → 3.34%),WOE 单调上升(-0.1555 → 0.9608),符合风控分箱预期。IV=0.1475,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 1 箱((-inf, 11.0])为坏率最高区间:坏率高达 9.53%,可作为拒绝或重点风控策略的阈值参考

cre_l90_used_limit_amt_min — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 21.640000343322754]561854082103.74%0.8430
2(21.640000343322754, 748.7099914550781]807675265506.81%0.2107
3(748.7099914550781, 8516.22998046875]201721835418189.01%-0.0934
4(8516.22998046875, 14597.384765625]2267199027712.22%-0.4336
5(14597.384765625, inf]1932163729515.27%-0.6919
2026-10-27T09:47:26.029931 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(3.74% → 15.27%),WOE 单调下降(0.8430 → -0.6919),符合风控分箱预期。IV=0.1336,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 5 箱((14597.384765625, inf])为坏率最高区间:坏率高达 15.27%,可作为拒绝或重点风控策略的阈值参考

cre_now_l60_used_amt_m_min — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.7799999713897705]282012545827439.73%-0.1775
2(1.7799999713897705, inf]986494574074.13%0.7402
2026-10-27T09:47:26.148004 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(9.73% → 4.13%),WOE 单调上升(-0.1775 → 0.7402),符合风控分箱预期。IV=0.1300,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 1 箱((-inf, 1.7799999713897705])为坏率最高区间:坏率高达 9.73%,可作为拒绝或重点风控策略的阈值参考

apprv_num_m1 — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]281522542527279.69%-0.1734
2(0.5, 1.5]535051072434.54%0.6394
3(1.5, inf]457843981803.93%0.7900
2026-10-27T09:47:26.254936 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(9.69% → 3.93%),WOE 单调上升(-0.1734 → 0.7900),符合风控分箱预期。IV=0.1224,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])为坏率最高区间:坏率高达 9.69%,可作为拒绝或重点风控策略的阈值参考

s_ps_od_days_max — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]2061018430218010.58%-0.2713
2(0.5, inf]17470165009705.55%0.4279
2026-10-27T09:47:26.362102 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(10.58% → 5.55%),WOE 单调上升(-0.2713 → 0.4279),符合风控分箱预期。IV=0.1150,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 1 箱((-inf, 0.5])为坏率最高区间:坏率高达 10.58%,可作为拒绝或重点风控策略的阈值参考

available_limit_amt — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.03499999921768904]1238510918146711.85%-0.3987
2(0.03499999921768904, 27316.265625]237712218315886.68%0.2309
3(27316.265625, inf]19241829954.94%0.5517
2026-10-27T09:47:26.471279 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(11.85% → 4.94%),WOE 单调上升(-0.3987 → 0.5517),符合风控分箱预期。IV=0.1036,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, 0.03499999921768904])为坏率最高区间:坏率高达 11.85%,可作为拒绝或重点风控策略的阈值参考

cre_now_l30_used_amt_d_min — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.0000500082969666]312112827329389.41%-0.0980
2(1.0000500082969666, 1.382200002670288]19331860733.78%0.8757
3(1.382200002670288, inf]27682687812.93%1.1395
2026-10-27T09:47:26.586679 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(9.41% → 2.93%),WOE 单调上升(-0.0980 → 1.1395),符合风控分箱预期。IV=0.1011,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, 1.0000500082969666])为坏率最高区间:坏率高达 9.41%,可作为拒绝或重点风控策略的阈值参考

max_od_days — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 5.5]313952916222337.11%0.1636
2(5.5, inf]6685576891713.72%-0.5670
2026-10-27T09:47:26.704686 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(7.11% → 13.72%),WOE 单调下降(0.1636 → -0.5670),符合风控分箱预期。IV=0.0920,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 2 箱((5.5, inf])为坏率最高区间:坏率高达 13.72%,可作为拒绝或重点风控策略的阈值参考

setl_prcp_ratio — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.035099999979138374]832579603654.38%0.6763
2(0.035099999979138374, inf]297552697027859.36%-0.1355
2026-10-27T09:47:26.816470 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(4.38% → 9.36%),WOE 单调下降(0.6763 → -0.1355),符合风控分箱预期。IV=0.0909,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 2 箱((0.035099999979138374, inf])为坏率最高区间:坏率高达 9.36%,可作为拒绝或重点风控策略的阈值参考

end_loan_to_cur_freq — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -22.5]299592716227979.34%-0.1327
2(-22.5, inf]812177683534.35%0.6854
2026-10-27T09:47:26.925745 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(9.34% → 4.35%),WOE 单调上升(-0.1327 → 0.6854),符合风控分箱预期。IV=0.0902,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, -22.5])为坏率最高区间:坏率高达 9.34%,可作为拒绝或重点风控策略的阈值参考

loan_os_prcp_credit_amount_m12 — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 16.18000030517578]316430581063.35%0.9564
2(16.18000030517578, 44.21000099182129]603356713626.00%0.3458
3(44.21000099182129, 77.875]144531322312308.51%-0.0308
4(77.875, 95.76499938964844]663560086279.45%-0.1458
5(95.76499938964844, inf]7775695182410.60%-0.2732
2026-10-27T09:47:27.046724 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(3.35% → 10.60%),WOE 单调下降(0.9564 → -0.2732),符合风控分箱预期。IV=0.0894,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((95.76499938964844, inf])为坏率最高区间:坏率高达 10.60%,可作为拒绝或重点风控策略的阈值参考

cre_now_l60_used_amt_d_min — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.000950038433075]279432520527389.80%-0.1106
2(1.000950038433075, 1.4549500346183777]235622351215.14%0.5858
3(1.4549500346183777, inf]388637151714.40%0.7481
2026-10-27T09:47:27.165332 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(9.80% → 4.40%),WOE 单调上升(-0.1106 → 0.7481),符合风控分箱预期。IV=0.0761,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 1 箱((-inf, 1.000950038433075])为坏率最高区间:坏率高达 9.80%,可作为拒绝或重点风控策略的阈值参考

cre_now_l30_limit_left_d_max — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.9999499917030334]721468873274.53%0.6418
2(0.9999499917030334, inf]306002779828029.16%-0.1110
2026-10-27T09:47:27.272949 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(4.53% → 9.16%),WOE 单调下降(0.6418 → -0.1110),符合风控分箱预期。IV=0.0708,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 2 箱((0.9999499917030334, inf])为坏率最高区间:坏率高达 9.16%,可作为拒绝或重点风控策略的阈值参考

shd_pre_repay_due_rate — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.868800014257431]1047399205535.28%0.4846
2(0.868800014257431, inf]271912461825739.46%-0.1439
2026-10-27T09:47:27.432013 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(5.28% → 9.46%),WOE 单调下降(0.4846 → -0.1439),符合风控分箱预期。IV=0.0693,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 2 箱((0.868800014257431, inf])为坏率最高区间:坏率高达 9.46%,可作为拒绝或重点风控策略的阈值参考

loan_os_prcp_gt0_m3_num_pc — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 68.16500091552734]335432521023.04%1.0563
2(68.16500091552734, inf]347063165930478.78%-0.0649
2026-10-27T09:47:27.536904 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(3.04% → 8.78%),WOE 单调下降(1.0563 → -0.0649),符合风控分箱预期。IV=0.0681,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 2 箱((68.16500091552734, inf])为坏率最高区间:坏率高达 8.78%,可作为拒绝或重点风控策略的阈值参考

loan_os_prcp_credit_amount_m6 — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 33.795000076293945]812777343934.84%0.5739
2(33.795000076293945, inf]299332717727569.21%-0.1171
2026-10-27T09:47:27.646136 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(4.84% → 9.21%),WOE 单调下降(0.5739 → -0.1171),符合风控分箱预期。IV=0.0668,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 2 箱((33.795000076293945, inf])为坏率最高区间:坏率高达 9.21%,可作为拒绝或重点风控策略的阈值参考

ps_int_amt_sum — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 209.51499938964844]317330021715.39%0.4594
2(209.51499938964844, 815.2900085449219]1035696177397.14%0.1600
3(815.2900085449219, 1891.7449951171875]1042996138167.82%0.0605
4(1891.7449951171875, 10533.93994140625]113161027610409.19%-0.1154
5(10533.93994140625, inf]2806242238413.69%-0.5642
2026-10-27T09:47:27.762304 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(5.39% → 13.69%),WOE 单调下降(0.4594 → -0.5642),符合风控分箱预期。IV=0.0558,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 5 箱((10533.93994140625, inf])为坏率最高区间:坏率高达 13.69%,可作为拒绝或重点风控策略的阈值参考

useratemore50maxcontimonsm12 — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]407638891874.59%0.6291
2(0.5, 2.5]428940352545.92%0.3597
3(2.5, inf]296952698727089.12%-0.1066
2026-10-27T09:47:27.881339 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(4.59% → 9.12%),WOE 单调下降(0.6291 → -0.1066),符合风控分箱预期。IV=0.0545,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 1 箱 WOE<0(风险高于平均)

第 3 箱((2.5, inf])为坏率最高区间:坏率高达 9.12%,可作为拒绝或重点风控策略的阈值参考

avg_tnr — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 3.5]24362340963.94%0.7876
2(3.5, 5.5]367234422306.26%0.2998
3(5.5, 6.5]266024721887.07%0.1704
4(6.5, 10.5]126711164510268.10%0.0233
5(10.5, inf]166411503116109.67%-0.1721
2026-10-27T09:47:27.998213 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(3.94% → 9.67%),WOE 单调下降(0.7876 → -0.1721),符合风控分箱预期。IV=0.0524,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 5 箱((10.5, inf])为坏率最高区间:坏率高达 9.67%,可作为拒绝或重点风控策略的阈值参考

cre_now_l60_used_amt_m_max — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 83.30500030517578]514849042444.74%0.5951
2(83.30500030517578, 2003.010009765625]240882208320058.32%-0.0063
3(2003.010009765625, inf]8829792890110.21%-0.2309
2026-10-27T09:47:28.112424 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(4.74% → 10.21%),WOE 单调下降(0.5951 → -0.2309),符合风控分箱预期。IV=0.0511,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((2003.010009765625, inf])为坏率最高区间:坏率高达 10.21%,可作为拒绝或重点风控策略的阈值参考

max_loan_os_prcp_m12 — 4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1747.2849731445312]410138602415.88%0.3679
2(1747.2849731445312, 13572.5849609375]252862332819587.74%0.0720
3(13572.5849609375, 20848.58984375]361932783419.42%-0.1426
4(20848.58984375, inf]5054444560912.05%-0.4180
2026-10-27T09:47:28.232529 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(5.88% → 12.05%),WOE 单调下降(0.3679 → -0.4180),符合风控分箱预期。IV=0.0455,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 2 箱 WOE<0(风险高于平均)

第 4 箱((20848.58984375, inf])为坏率最高区间:坏率高达 12.05%,可作为拒绝或重点风控策略的阈值参考

setl_rem_prcp_paym_m3_num — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]320942924028548.89%-0.0789
2(0.5, inf]596656712954.94%0.5504
2026-10-27T09:47:28.353776 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(8.89% → 4.94%),WOE 单调上升(-0.0789 → 0.5504),符合风控分箱预期。IV=0.0433,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])为坏率最高区间:坏率高达 8.89%,可作为拒绝或重点风控策略的阈值参考

orig_prcp_avg_m12 — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1031.9722290039062]11498107357636.64%0.2381
2(1031.9722290039062, 1871.875]692564055207.51%0.1051
3(1871.875, 3819.857177734375]872779937348.41%-0.0181
4(3819.857177734375, 13787.5]890980418689.74%-0.1798
5(13787.5, inf]2021175626513.11%-0.5149
2026-10-27T09:47:28.477112 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(6.64% → 13.11%),WOE 单调下降(0.2381 → -0.5149),符合风控分箱预期。IV=0.0431,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((13787.5, inf])为坏率最高区间:坏率高达 13.11%,可作为拒绝或重点风控策略的阈值参考

useratemore70_m6_num — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]10817101466716.20%0.3103
2(0.5, 1.5]443541153207.22%0.1484
3(1.5, 4.5]144121310413089.08%-0.1013
4(4.5, 5.5]346831323369.69%-0.1734
5(5.5, inf]4928441451410.43%-0.2554
2026-10-27T09:47:28.603138 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(6.20% → 10.43%),WOE 单调下降(0.3103 → -0.2554),符合风控分箱预期。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 5 箱((5.5, inf])为坏率最高区间:坏率高达 10.43%,可作为拒绝或重点风控策略的阈值参考

useratemore70monthsubcurmonth — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4.5]254292310223279.15%-0.1104
2(4.5, 9.5]603755644737.83%0.0593
3(9.5, inf]659462453495.29%0.4787
2026-10-27T09:47:28.782774 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(9.15% → 5.29%),WOE 单调上升(-0.1104 → 0.4787),符合风控分箱预期。IV=0.0416,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 1 箱((-inf, 4.5])为坏率最高区间:坏率高达 9.15%,可作为拒绝或重点风控策略的阈值参考

suc_apprv_num_m6 — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1.5]220111992520869.48%-0.1492
2(1.5, 3.5]781272405727.32%0.1323
3(3.5, inf]825777654925.96%0.3530
2026-10-27T09:47:28.896847 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(9.48% → 5.96%),WOE 单调上升(-0.1492 → 0.3530),符合风控分箱预期。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 2 箱 WOE<0(风险高于平均)

第 1 箱((-inf, 1.5])为坏率最高区间:坏率高达 9.48%,可作为拒绝或重点风控策略的阈值参考

shd_pre_repay_amt_rate — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.6238499879837036]12637118148236.51%0.2584
2(0.6238499879837036, 0.7506499886512756]863179446877.96%0.0421
3(0.7506499886512756, inf]167911515216399.76%-0.1816
2026-10-27T09:47:29.007325 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(6.51% → 9.76%),WOE 单调下降(0.2584 → -0.1816),符合风控分箱预期。IV=0.0360,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 1 箱 WOE<0(风险高于平均)

第 3 箱((0.7506499886512756, inf])为坏率最高区间:坏率高达 9.76%,可作为拒绝或重点风控策略的阈值参考

useratemore90maxcontimonsm12 — 5 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]929887235756.18%0.3136
2(0.5, 1.5]690363495548.03%0.0332
3(1.5, 4.5]147501346112898.74%-0.0598
4(4.5, 7.5]465241884649.97%-0.2056
5(7.5, inf]2457219026710.87%-0.3013
2026-10-27T09:47:29.127125 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(6.18% → 10.87%),WOE 单调下降(0.3136 → -0.3013),符合风控分箱预期。IV=0.0350,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((7.5, inf])为坏率最高区间:坏率高达 10.87%,可作为拒绝或重点风控策略的阈值参考

ps_int_amt_avg — 4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 4.333299875259399]408338572265.54%0.4312
2(4.333299875259399, 13.080849647521973]12227113149137.47%0.1111
3(13.080849647521973, 38.24515151977539]143511310612458.68%-0.0520
4(38.24515151977539, inf]7419665376610.32%-0.2443
2026-10-27T09:47:29.246882 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(5.54% → 10.32%),WOE 单调下降(0.4312 → -0.2443),符合风控分箱预期。IV=0.0344,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 2 箱 WOE<0(风险高于平均)

第 4 箱((38.24515151977539, inf])为坏率最高区间:坏率高达 10.32%,可作为拒绝或重点风控策略的阈值参考

repay_amount_m6_all_shd_amt_pc — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 93.98500061035156]132571194613119.89%-0.1961
2(93.98500061035156, 99.48500061035156]197431819315507.85%0.0571
3(99.48500061035156, inf]506047722885.69%0.4018
2026-10-27T09:47:29.360719 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(9.89% → 5.69%),WOE 单调上升(-0.1961 → 0.4018),符合风控分箱预期。IV=0.0344,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 1 箱((-inf, 93.98500061035156])为坏率最高区间:坏率高达 9.89%,可作为拒绝或重点风控策略的阈值参考

cre_l30_l90_used_amt_m_min — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.6299999952316284]298562718526718.95%-0.0853
2(0.6299999952316284, inf]820977304795.84%0.3756
2026-10-27T09:47:29.468041 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(8.95% → 5.84%),WOE 单调上升(-0.0853 → 0.3756),符合风控分箱预期。IV=0.0320,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, 0.6299999952316284])为坏率最高区间:坏率高达 8.95%,可作为拒绝或重点风控策略的阈值参考

cre_now_p1m_used_rate_m — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -4.999999873689376e-05]728868704185.74%0.3962
2(-4.999999873689376e-05, inf]304462773527118.90%-0.0779
2026-10-27T09:47:29.574003 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(5.74% → 8.90%),WOE 单调下降(0.3962 → -0.0779),符合风控分箱预期。IV=0.0308,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 2 箱((-4.999999873689376e-05, inf])为坏率最高区间:坏率高达 8.90%,可作为拒绝或重点风控策略的阈值参考

useratemore90_m3_num — 4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]232662150117657.59%0.0942
2(0.5, 1.5]611756464717.70%0.0781
3(1.5, 2.5]397936233568.95%-0.0856
4(2.5, inf]4698414155711.86%-0.3996
2026-10-27T09:47:29.683894 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(7.59% → 11.86%),WOE 单调下降(0.0942 → -0.3996),符合风控分箱预期。IV=0.0302,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 2 箱 WOE<0(风险高于平均)

第 4 箱((2.5, inf])为坏率最高区间:坏率高达 11.86%,可作为拒绝或重点风控策略的阈值参考

min_apply_amt — 4 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 1135.0]3388301037811.16%-0.3311
2(1135.0, 4180.0]179831641315708.73%-0.0589
3(4180.0, 7300.0]922184907317.93%0.0463
4(7300.0, inf]748870174716.29%0.2953
2026-10-27T09:47:29.797837 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(11.16% → 6.29%),WOE 单调上升(-0.3311 → 0.2953),符合风控分箱预期。IV=0.0286,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 2 箱 WOE<0(风险高于平均)

第 1 箱((-inf, 1135.0])为坏率最高区间:坏率高达 11.16%,可作为拒绝或重点风控策略的阈值参考

cre_now_l30_limit_left_d_min — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.997050017118454]6136544369311.29%-0.3306
2(0.997050017118454, inf]303392797523647.79%0.0793
2026-10-27T09:47:29.905122 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(11.29% → 7.79%),WOE 单调上升(-0.3306 → 0.0793),符合风控分箱预期。IV=0.0262,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, 0.997050017118454])为坏率最高区间:坏率高达 11.29%,可作为拒绝或重点风控策略的阈值参考

loan_os_prcp_gt0_m12_num_pc — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 57.73500061035156]329831381604.85%0.5705
2(57.73500061035156, 72.08000183105469]263124172148.13%0.0186
3(72.08000183105469, inf]321312935627758.64%-0.0469
2026-10-27T09:47:30.017982 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(4.85% → 8.64%),WOE 单调下降(0.5705 → -0.0469),符合风控分箱预期。IV=0.0242,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 1 箱 WOE<0(风险高于平均)

第 3 箱((72.08000183105469, inf])为坏率最高区间:坏率高达 8.64%,可作为拒绝或重点风控策略的阈值参考

cre_l30_l90_usedrate_m_max — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.025150000117719173]11879110807996.73%0.2240
2(0.025150000117719173, 0.0799499973654747]331030532577.76%0.0693
3(0.0799499973654747, inf]228762078220949.15%-0.1105
2026-10-27T09:47:30.193533 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(6.73% → 9.15%),WOE 单调下降(0.2240 → -0.1105),符合风控分箱预期。IV=0.0224,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 1 箱 WOE<0(风险高于平均)

第 3 箱((0.0799499973654747, inf])为坏率最高区间:坏率高达 9.15%,可作为拒绝或重点风控策略的阈值参考

cre_now_l60_limit_left_d_min — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.9717499911785126]8039718685310.61%-0.2558
2(0.9717499911785126, inf]278622569321697.78%0.0850
2026-10-27T09:47:30.305560 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(10.61% → 7.78%),WOE 单调上升(-0.2558 → 0.0850),符合风控分箱预期。IV=0.0217,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, 0.9717499911785126])为坏率最高区间:坏率高达 10.61%,可作为拒绝或重点风控策略的阈值参考

min_tnr — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 5.0]229492126016897.36%0.1267
2(5.0, inf]151311367014619.66%-0.1699
2026-10-27T09:47:30.410862 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(7.36% → 9.66%),WOE 单调下降(0.1267 → -0.1699),符合风控分箱预期。IV=0.0215,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 2 箱((5.0, inf])为坏率最高区间:坏率高达 9.66%,可作为拒绝或重点风控策略的阈值参考

last_clear_num_m1 — 3 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, 0.5]301612750926528.79%-0.0667
2(0.5, 1.5]491645943226.55%0.2520
3(1.5, inf]300328271765.86%0.3705
2026-10-27T09:47:30.512082 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(8.79% → 5.86%),WOE 单调上升(-0.0667 → 0.3705),符合风控分箱预期。IV=0.0203,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 1 箱((-inf, 0.5])为坏率最高区间:坏率高达 8.79%,可作为拒绝或重点风控策略的阈值参考

setl_rem_prcp_paym_last_month_lo — 2 箱
分组号分栏TotalGoodBadBad RateWOE
1(-inf, -4999.0]289927571424.90%0.5604
2(-4999.0, inf]351613215430078.55%-0.0361
2026-10-27T09:47:30.618918 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(4.90% → 8.55%),WOE 单调下降(0.5604 → -0.0361),符合风控分箱预期。IV=0.0202,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 2 箱((-4999.0, inf])为坏率最高区间:坏率高达 8.55%,可作为拒绝或重点风控策略的阈值参考

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)。当前流程细/粗分箱由同一自动分箱算法直接产出。

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 125,成功分箱 125,入模 45

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
ps_perd_no—N0.1442PASSPASSPASS—0.3215PSI>0.25
due_day_my—Y19.9031PASSPASSPASSPASS0.0005
age—N0.0040PASSPASSFAIL——IV<0.02
max_int_rat—N0.0079PASSPASSFAIL——IV<0.02
min_int_rat—N0.0016PASSPASSFAIL——IV<0.02
avg_tnr—Y0.0524PASSPASSPASSPASS0.0410
min_tnr—Y0.0215PASSPASSPASSPASS0.0021
sum_apply_amt—N0.0023PASSPASSFAIL——IV<0.02
min_apply_amt—Y0.0286PASSPASSPASSPASS0.0266
min_apprv_amt—N0.0041PASSPASSFAIL——IV<0.02
apprv_amt_day—Y0.1475PASSPASSPASSPASS0.1320
end_loan_to_rep_loan_freq—N0.0027PASSPASSFAIL——IV<0.02
end_loan_to_cur_freq—Y0.0902PASSPASSPASSPASS0.1153
f_rep_loan_to_f_apply_dt_freq—N0.0010PASSPASSFAIL——IV<0.02
ps_prcp_amt_mon—N0.0267PASSPASSPASS—0.3348PSI>0.25
ps_int_amt_mon—N0.0946PASSPASSPASS—0.3037PSI>0.25
ps_int_amt_sum—Y0.0558PASSPASSPASSPASS0.2234
ps_prcp_amt_avg—N0.0043PASSPASSFAIL——IV<0.02
ps_int_amt_avg—Y0.0344PASSPASSPASSPASS0.1191
setl_prcp_mon_ratio—N0.0121PASSPASSFAIL——IV<0.02
acc_ps_perd_no—N0.0987PASSPASSPASS—0.4588PSI>0.25
max_od_days—Y0.0920PASSPASSPASSPASS0.2317
m_s_ps_instm_amt—N0.0363PASSPASSPASS—0.2758PSI>0.25
f_ps_od_days_max—Y0.1798PASSPASSPASSPASS0.1412
s_ps_od_days_max—Y0.1150PASSPASSPASSPASS0.1764
t_ps_od_days_max—N0.0168PASSPASSFAIL——IV<0.02
ovd_days_d3_num_m1—N0.1789PASSFAIL———集中度>0.9
orig_prcp_avg_m3—N0.1156PASSPASSPASS—0.2544PSI>0.25
orig_prcp_avg_m12—Y0.0431PASSPASSPASSPASS0.0373
suc_apply_trench_num—N0.0091PASSFAIL———集中度>0.9
suc_apply_top_pc_code_num—N0.0022PASSPASSFAIL——IV<0.02
suc_apprv_num_m1—Y0.1594PASSPASSPASSPASS0.1386
apprv_num_m1—Y0.1224PASSPASSPASSPASS0.1073
suc_apply_num_m3—N0.0000PASSFAIL———集中度>0.9
apprv_num_m3—N0.0951PASSPASSPASSFAIL0.2387与 apprv_num_m1 相关>0.7
suc_apprv_num_m6—Y0.0404PASSPASSPASSPASS0.0130
apply_num_m12—N0.0007PASSPASSFAIL——IV<0.02
last_apply_od_num_m12—N0.0067PASSFAIL———集中度>0.9
last_apply_freq—N0.0169PASSPASSFAIL——IV<0.02
dec_loan_num—N0.0028PASSFAIL———集中度>0.9
bh_cur_od_num—N0.0064PASSPASSFAIL——IV<0.02
hj_cur_od_num—N0.0005PASSFAIL———集中度>0.9
setl_prcp_ratio—Y0.0909PASSPASSPASSPASS0.2377
last_loan_setl_tnr—N0.1487PASSPASSPASS—0.2771PSI>0.25
last_od_num_m6—N0.1092PASSPASSPASS—0.5913PSI>0.25
last_od_to_cur_freq—N0.1995PASSPASSPASS—0.7926PSI>0.25
last_clear_num_m1—Y0.0203PASSPASSPASSPASS0.0046
last_clear_num_m3—N0.0305PASSPASSPASSFAIL0.0139与 suc_apprv_num_m6 相关>0.7
grade_limit_amount—N0.0103PASSPASSFAIL——IV<0.02
exposure—N0.0033PASSPASSFAIL——IV<0.02
available_limit_amt—Y0.1036PASSPASSPASSPASS0.0068
account_age—N0.0131PASSPASSFAIL——IV<0.02
account_age_loan_day—N0.0048PASSPASSFAIL——IV<0.02
max_od_due—N0.0638PASSPASSPASS—0.4345PSI>0.25
odgt0_m12_mob_pc—N0.0442PASSPASSPASS—0.9380PSI>0.25
odgt0_m3_num—N0.0227PASSPASSPASS—1.0906PSI>0.25
odgt0_m12_num—N0.0464PASSPASSPASS—0.9376PSI>0.25
odgt0_m12_subcurmonth—N0.0635PASSPASSPASS—1.2145PSI>0.25
tothismonthodcontiaddnumm12—N0.0187PASSPASSFAIL——IV<0.02
tothismonthodcontireducenumm12—N0.0209PASSPASSPASS—0.7206PSI>0.25
od_amt_m6_loan_os_prcp_m6_pc—N0.0804PASSFAIL———集中度>0.9
od_amt_m12_loan_os_prcp_m12_pc—N0.0658PASSFAIL———集中度>0.9
avg_od_amt_m6_orig_prcp_pc—N0.0815PASSFAIL———集中度>0.9
max_loan_os_prcp_m12—Y0.0455PASSPASSPASSPASS0.0637
loan_os_prcp_gt0_m3_num_pc—Y0.0681PASSPASSPASSPASS0.0255
loan_os_prcp_gt0_m12_num_pc—Y0.0242PASSPASSPASSPASS0.0334
repay_amount_m3_loan_os_prcp_pc—N0.0283PASSPASSPASS—0.3012PSI>0.25
repay_amount_m6_loan_os_prcp_pc—N0.0070PASSPASSFAIL——IV<0.02
repay_amount_all_shd_amt_pc—N0.0033PASSPASSFAIL——IV<0.02
repay_amount_m6_all_shd_amt_pc—Y0.0344PASSPASSPASSPASS0.0382
setl_rem_prcp_paym_last_month_lo—Y0.0202PASSPASSPASSPASS0.1082
setl_rem_prcp_paym_m12_loan_os_p—N0.0039PASSPASSFAIL——IV<0.02
setl_rem_prcp_paym_m3_num—Y0.0433PASSPASSPASSPASS0.0163
setl_rem_prcp_paym_m12_num—N0.0047PASSPASSFAIL——IV<0.02
loan_os_prcp_credit_amount_m3—N0.0068PASSPASSFAIL——IV<0.02
loan_os_prcp_credit_amount_m6—Y0.0668PASSPASSPASSPASS0.0356
loan_os_prcp_credit_amount_m12—Y0.0894PASSPASSPASSPASS0.0786
useratemore50maxcontimonsm12—Y0.0545PASSPASSPASSPASS0.1876
useratemore70_m6_num—Y0.0429PASSPASSPASSPASS0.0784
useratemore70monthsubcurmonth—Y0.0416PASSPASSPASSPASS0.0847
useratemore90_m3_num—Y0.0302PASSPASSPASSPASS0.0027
useratemore90maxcontimonsm12—Y0.0350PASSPASSPASSPASS0.1030
shd_max_dpd—N0.0690PASSPASSPASS—0.3133PSI>0.25
shd_min_dpd—N0.0027PASSPASSFAIL——IV<0.02
shd_pre_repay_cnt—N0.0181PASSPASSFAIL——IV<0.02
shd_perd_max—N0.0161PASSFAIL———集中度>0.9
shd_instm_amt_sum—N0.0258PASSPASS———唯一率过高(ID列)
shd_l30_max_dpd—N0.0576PASSPASSPASS—1.1956PSI>0.25
shd_l60_min_dpd—N0.0022PASSPASSFAIL——IV<0.02
shd_l90_ps_due_cnt—N0.0275PASSPASSPASS—0.3391PSI>0.25
shd_l180_max_dpd—N0.0043PASSPASSFAIL——IV<0.02
shd_l180_instm_amt_sum—N0.0229PASSPASS———唯一率过高(ID列)
shd_remain_actual_cnt—N0.0129PASSPASSFAIL——IV<0.02
shd_pre_repay_cnt_rate—N0.0124PASSPASSFAIL——IV<0.02
shd_pre_repay_amt_rate—Y0.0360PASSPASSPASSPASS0.0366
shd_pre_repay_30d_amt_rate—N0.0003PASSPASSFAIL——IV<0.02
shd_pre_repay_due_rate—Y0.0693PASSPASSPASSPASS0.2464
shd_l30_pre_repay_due_rate—N0.0506PASSPASSPASS—0.8903PSI>0.25
shd_remain_actual_rate—N0.0095PASSPASSFAIL——IV<0.02
cre_usedrate_u30_daydelta_min—N0.0116PASSPASSFAIL——IV<0.02
cre_usedrate_u30_daydelta_max—N0.0468PASSPASSPASS—0.2756PSI>0.25
cre_l30_limit_left_amt_min—N0.0459PASSPASSPASSFAIL0.0010与 available_limit_amt 相关>0.7
cre_l60_used_limit_rate_max—N0.0614PASSPASSPASSFAIL0.0130与 cre_l60_used_limit_rate_min 相关>0.7
cre_l60_used_limit_rate_min—Y0.1478PASSPASSPASSPASS0.0410
cre_l90_used_limit_amt_min—Y0.1336PASSPASSPASSPASS0.0680
cre_now_p1m_used_rate_d—N0.0140PASSPASSFAIL——IV<0.02
cre_now_p3m_used_rate_d—N0.0161PASSPASSFAIL——IV<0.02
cre_now_p1m_used_rate_m—Y0.0308PASSPASSPASSPASS0.0095
cre_now_p2m_used_rate_m—N0.0190PASSPASSFAIL——IV<0.02
cre_now_l30_used_amt_d_max—N0.0151PASSPASSFAIL——IV<0.02
cre_now_l30_used_amt_d_min—Y0.1011PASSPASSPASSPASS0.0345
cre_now_l30_limit_left_d_max—Y0.0708PASSPASSPASSPASS0.0105
cre_now_l30_limit_left_d_min—Y0.0262PASSPASSPASSPASS0.0026
cre_now_l60_used_amt_d_min—Y0.0761PASSPASSPASSPASS0.0585
cre_now_l60_limit_left_d_max—N0.0088PASSPASSFAIL——IV<0.02
cre_now_l60_limit_left_d_min—Y0.0217PASSPASSPASSPASS0.0108
cre_now_l60_used_amt_m_max—Y0.0511PASSPASSPASSPASS0.2386
cre_now_l60_used_amt_m_min—Y0.1300PASSPASSPASSPASS0.0710
cre_now_l60_usedrate_m_max—N0.0213PASSPASSPASSFAIL0.1554与 cre_l30_l90_usedrate_m_max 相关>0.7
cre_now_l90_limit_left_d_min—N0.0204PASSPASSPASSFAIL0.0181与 cre_now_l60_limit_left_d_min 相关>0.7
cre_l30_l90_used_amt_d_max—N0.0705PASSPASSPASS—0.3024PSI>0.25
cre_l30_l90_used_amt_m_max—N0.0868PASSPASSPASS—0.3024PSI>0.25
cre_l30_l90_used_amt_m_min—Y0.0320PASSPASSPASSPASS0.0504
cre_l30_l90_usedrate_m_max—Y0.0224PASSPASSPASSPASS0.1840
cre_l30_l90_usedrate_m_min—N0.0034PASSPASSFAIL——IV<0.02

06 回归结果

硬门槛:P≥0.1 / VIF>4.0 迭代剔除,共 40 个变量被剔除(41 轮):cre_now_l60_limit_left_d_min、cre_now_l30_limit_left_d_min、cre_l30_l90_usedrate_m_max、min_tnr、setl_rem_prcp_paym_last_month_lo、last_clear_num_m1、cre_now_p1m_used_rate_m、useratemore90_m3_num、loan_os_prcp_gt0_m12_num_pc、useratemore90maxcontimonsm12、cre_l30_l90_used_amt_m_min、useratemore70monthsubcurmonth、min_apply_amt、ps_int_amt_avg、repay_amount_m6_all_shd_amt_pc、shd_pre_repay_amt_rate、orig_prcp_avg_m12、useratemore70_m6_num、suc_apprv_num_m6、max_loan_os_prcp_m12、useratemore50maxcontimonsm12、setl_rem_prcp_paym_m3_num、cre_now_l60_used_amt_m_max、cre_now_l60_used_amt_d_min、cre_now_l30_limit_left_d_max、ps_int_amt_sum、avg_tnr、loan_os_prcp_credit_amount_m6、loan_os_prcp_gt0_m3_num_pc、setl_prcp_ratio、apprv_amt_day、shd_pre_repay_due_rate、max_od_days、end_loan_to_cur_freq、loan_os_prcp_credit_amount_m12、apprv_num_m1、available_limit_amt、s_ps_od_days_max、cre_now_l60_used_amt_m_min、cre_now_l30_used_amt_d_min。

ParameterLabelEstimate(无惩罚)Standard ErrorWald Chi-SquarePr > ChiSqStandardized Estimate
InterceptIntercept-0.132713.52420.000.9922—
WOE_due_day_mydue_day_my-1.34751.54470.760.3830-7.3878
WOE_f_ps_od_days_maxf_ps_od_days_max-0.01394.72230.000.9976-0.0061
WOE_suc_apprv_num_m1suc_apprv_num_m1-0.01354.51380.000.9976-0.0065
WOE_cre_l60_used_limit_rate_mincre_l60_used_limit_rate_min-0.01186.70780.000.9986-0.0051
WOE_cre_l90_used_limit_amt_mincre_l90_used_limit_amt_min-0.01047.25590.000.9989-0.0040

注:Estimate 列为无惩罚(精确推断)拟合;生产系数保持 L2 正则(model_results.json / score.py 口径),两套估计并列展示。标准误基于无惩罚似然观测信息矩阵。

07 VIF

VIF

ParameterVIF
due_day_my1.02603
f_ps_od_days_max1.01813
suc_apprv_num_m11.13120
cre_l60_used_limit_rate_min1.93071
cre_l90_used_limit_amt_min1.84497

相关系数(WOE 转换后,开发样本)

due_day_myf_ps_od_days_maxsuc_apprv_num_m1cre_l60_used_limit_rate_mincre_l90_used_limit_amt_min
due_day_my1.0000.1130.0910.0950.095
f_ps_od_days_max0.1131.0000.0770.0530.033
suc_apprv_num_m10.0910.0771.0000.3280.256
cre_l60_used_limit_rate_min0.0950.0530.3281.0000.675
cre_l90_used_limit_amt_min0.0950.0330.2560.6751.000

08 最终评分卡

标准分600.0
标准ODD20.0
PDO20.0
变量分栏WOECoefficientScorecard Points
due_day_my(-inf, 30.5]8.7486-1.3214437
(30.5, inf]-11.1544-1.3214-322
f_ps_od_days_max(-inf, 0.5]-0.3309-0.0137103
(0.5, inf]0.5516-0.0137104
suc_apprv_num_m1(-inf, 0.5]-0.1580-0.0133103
(0.5, 1.5]0.7478-0.0133104
(1.5, inf]1.3409-0.0133104
cre_l60_used_limit_rate_min(-inf, 0.019700000062584877]1.0570-0.0116104
(0.019700000062584877, 0.09174999967217445]0.4228-0.0116104
(0.09174999967217445, 0.2787500023841858]0.1741-0.0116104
(0.2787500023841858, 0.9161500036716461]-0.1181-0.0116103
(0.9161500036716461, inf]-0.5111-0.0116103
cre_l90_used_limit_amt_min(-inf, 21.640000343322754]0.8430-0.0102104
(21.640000343322754, 748.7099914550781]0.2107-0.0102104
(748.7099914550781, 8516.22998046875]-0.0934-0.0102103
(8516.22998046875, 14597.384765625]-0.4336-0.0102103
(14597.384765625, inf]-0.6919-0.0102103

09 模型表现

模型区分度评估(按样本)

模型样本KSAUCGINI
LR + WOE 评分卡开发样本1.00001.00001.0000
LR + WOE 评分卡测试样本1.00001.00001.0000
LR + WOE 评分卡训练+测试样本1.00001.00001.0000
LR + WOE 评分卡跨时间验证样本1.00001.00001.0000

LR + WOE 评分卡 — 4 样本评估表(score_cut 粒度)

开发样本(KS 0.0362,AUC 0.5146)

cutscntbadgoodbad_ratepopulation_pctcum_bad_ratecum_good_rate
[92, 851)4035315188478.09%10.59%100.00%2.53%
[851, 851)7413074130.00%19.46%100.00%23.75%
[851, 851)3998039980.00%10.50%100.00%35.19%
[851, 851)3700037000.00%9.71%100.00%45.78%
[851, 851)4827048270.00%12.67%100.00%59.60%
[851, 851)3231032310.00%8.48%100.00%68.85%
[851, 852)3310033100.00%8.69%100.00%78.32%
[852, 852)4570045700.00%12.00%100.00%91.40%
[852, 852)3004030040.00%7.89%100.00%100.00%
2026-10-27T09:47:30.756337 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-10-27T09:47:30.882752 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

测试样本(KS 0.0325,AUC 0.5179)

cutscntbadgoodbad_ratepopulation_pctcum_bad_ratecum_good_rate
[92, 851)2653210055379.16%10.45%100.00%2.37%
[851, 851)4966049660.00%19.56%100.00%23.70%
[851, 851)2715027150.00%10.69%100.00%35.35%
[851, 851)2472024720.00%9.74%100.00%45.97%
[851, 851)3200032000.00%12.60%100.00%59.71%
[851, 851)2172021720.00%8.55%100.00%69.03%
[851, 852)2369023690.00%9.33%100.00%79.20%
[852, 852)2340023400.00%9.22%100.00%89.25%
[852, 852)2504025040.00%9.86%100.00%100.00%
2026-10-27T09:47:31.011326 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-10-27T09:47:31.132983 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

训练+测试样本(KS 0.0318,AUC 0.5159)

cutscntbadgoodbad_ratepopulation_pctcum_bad_ratecum_good_rate
[92, 851)66885251143778.51%10.54%100.00%2.47%
[851, 851)123790123790.00%19.50%100.00%23.73%
[851, 851)6713067130.00%10.58%100.00%35.26%
[851, 851)6172061720.00%9.72%100.00%45.86%
[851, 851)8027080270.00%12.65%100.00%59.64%
[851, 851)5403054030.00%8.51%100.00%68.92%
[851, 852)5441054410.00%8.57%100.00%78.26%
[852, 852)6316063160.00%9.95%100.00%89.11%
[852, 852)6340063400.00%9.99%100.00%100.00%
2026-10-27T09:47:31.259867 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-10-27T09:47:31.444205 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

跨时间验证样本(KS 0.0307,AUC 0.5201)

cutscntbadgoodbad_ratepopulation_pctcum_bad_ratecum_good_rate
[92, 851)39102432147862.20%11.26%100.00%4.58%
[851, 851)5809058090.00%16.73%100.00%22.56%
[851, 851)1105011050.00%3.18%100.00%25.98%
[851, 851)3779037790.00%10.88%100.00%37.68%
[851, 851)3210032100.00%9.24%100.00%47.62%
[851, 851)3396033960.00%9.78%100.00%58.14%
[851, 851)3538035380.00%10.19%100.00%69.09%
[851, 852)3070030700.00%8.84%100.00%78.60%
[852, 852)3970039700.00%11.43%100.00%90.89%
[852, 852)2943029430.00%8.47%100.00%100.00%
2026-10-27T09:47:31.573251 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-10-27T09:47:31.694890 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 100.0% 的坏客户,策略切分空间很大。

① 样本与分箱结构:共 34,730 样本 / 2,432 坏样本,整体坏率 7.00%,好坏比约 13.3:1;10 档等频分数分栏,每箱约 10%,个别箱占比偏薄/偏厚(3.2%~16.7%)。

② 排序性与单调性:逐箱 Bad Rate 62.20% → … → 0.00%(低分 → 高分),无倒挂。

③ 区分度与坏客户捕获:最低 20% 样本捕获全量 100.0% 的坏客户,最低 50% 捕获 100.0%;最高 20% 样本仅含 0.0% 的坏客户,其中风险最高一箱单独占 100.0%。

(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 表按分数等频分 10 档(低分 → 高分):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。

10 KS和GINI

LR + WOE 评分卡

样本KSGINIAUC
开发样本1.00001.00001.0000
测试样本1.00001.00001.0000
训练+测试样本1.00001.00001.0000
跨时间验证样本1.00001.00001.0000

开发样本

2026-10-27T09:47:31.781467 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-10-27T09:47:31.850232 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

测试样本

2026-10-27T09:47:31.917475 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-10-27T09:47:31.986807 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

训练+测试样本

2026-10-27T09:47:32.053837 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-10-27T09:47:32.122133 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

跨时间验证样本

2026-10-27T09:47:32.189538 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-10-27T09:47:32.257275 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oot验证集训练集与psi验证集
LR + WOE 评分卡0.00020.0358—0.0392

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,0.0]7.89%10.54%0.00770.0077
(0.0,0.0]12.00%16.17%0.01240.0201
(0.0,0.0]8.69%8.57%0.00000.0201
(0.0,0.0]8.48%6.90%0.00320.0234
(0.0,0.0]12.67%9.33%0.01020.0336
(0.0,0.0]9.71%9.31%0.00020.0338
(0.0,0.0]10.50%11.19%0.00040.0343
(0.0,0.0]0.00%0.00%0.00000.0343
(0.0,0.0]19.46%18.01%0.00110.0354
(0.0,inf]10.59%9.98%0.00040.0358

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,0.0]7.89%9.63%0.00350.0035
(0.0,0.0]12.00%15.82%0.01060.0140
(0.0,0.0]8.69%9.31%0.00040.0144
(0.0,0.0]8.48%6.82%0.00360.0181
(0.0,0.0]12.67%8.89%0.01340.0315
(0.0,0.0]9.71%9.21%0.00030.0317
(0.0,0.0]10.50%10.78%0.00010.0318
(0.0,0.0]0.00%0.00%0.00000.0318
(0.0,0.0]19.46%16.88%0.00370.0355
(0.0,inf]10.59%12.66%0.00370.0392
2026-10-27T09:47:32.334288 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

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, inf]30047.89%19767.78%49807.85%365910.54%16829.63%
(0.0, 0.0]457012.00%310612.23%767612.09%561616.17%276415.82%
(0.0, 0.0]33108.69%21318.39%54418.57%29788.57%16269.31%
(0.0, 0.0]32318.48%21728.55%54038.51%23986.90%11926.82%
(0.0, 0.0]482712.67%320012.60%802712.65%32409.33%15548.89%
(0.0, 0.0]37009.71%24729.74%61729.72%32329.31%16109.21%
(0.0, 0.0]399810.50%271510.69%671310.58%388811.19%188410.78%
(0.0, 0.0]00.00%00.00%00.00%00.00%00.00%
(0.0, 0.0]741319.46%496619.56%1237919.50%625418.01%294916.88%
(0.0, inf]403510.59%265310.45%668810.54%34659.98%221212.66%
sum38088100.00%25391100.00%63479100.00%34730100.00%17473100.00%

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
shd_pre_repay_due_rate0.2464关注
cre_now_l60_used_amt_m_max0.2386关注
setl_prcp_ratio0.2377关注
max_od_days0.2317关注
ps_int_amt_sum0.2234关注
useratemore50maxcontimonsm120.1876关注
cre_l30_l90_usedrate_m_max0.1840关注
s_ps_od_days_max0.1764关注
f_ps_od_days_max0.1412关注
suc_apprv_num_m10.1386关注
apprv_amt_day0.1320关注
ps_int_amt_avg0.1191关注
end_loan_to_cur_freq0.1153关注
setl_rem_prcp_paym_last_month_lo0.1082关注
apprv_num_m10.1073关注
useratemore90maxcontimonsm120.1030关注
useratemore70monthsubcurmonth0.0847稳定
loan_os_prcp_credit_amount_m120.0786稳定
useratemore70_m6_num0.0784稳定
cre_now_l60_used_amt_m_min0.0710稳定
cre_l90_used_limit_amt_min0.0680稳定
max_loan_os_prcp_m120.0637稳定
cre_now_l60_used_amt_d_min0.0585稳定
cre_l30_l90_used_amt_m_min0.0504稳定
avg_tnr0.0410稳定
cre_l60_used_limit_rate_min0.0410稳定
repay_amount_m6_all_shd_amt_pc0.0382稳定
orig_prcp_avg_m120.0373稳定
shd_pre_repay_amt_rate0.0366稳定
loan_os_prcp_credit_amount_m60.0356稳定
cre_now_l30_used_amt_d_min0.0345稳定
loan_os_prcp_gt0_m12_num_pc0.0334稳定
min_apply_amt0.0266稳定
loan_os_prcp_gt0_m3_num_pc0.0255稳定
setl_rem_prcp_paym_m3_num0.0163稳定
suc_apprv_num_m60.0130稳定
cre_now_l60_limit_left_d_min0.0108稳定
cre_now_l30_limit_left_d_max0.0105稳定
cre_now_p1m_used_rate_m0.0095稳定
available_limit_amt0.0068稳定
last_clear_num_m10.0046稳定
useratemore90_m3_num0.0027稳定
cre_now_l30_limit_left_d_min0.0026稳定
min_tnr0.0021稳定
due_day_my0.0005稳定
due_day_my — PSI 0.0005
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 30.5]91.73%91.12%-0.60%0.0000
(30.5, inf]8.27%8.88%0.60%0.0004
f_ps_od_days_max — PSI 0.1412
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]53.63%71.67%18.04%0.0523
(0.5, inf]46.37%28.33%-18.04%0.0889
suc_apprv_num_m1 — PSI 0.1386
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]79.23%62.50%-16.73%0.0397
(0.5, 1.5]12.57%22.98%10.41%0.0628
(1.5, inf]8.20%14.52%6.32%0.0361
cre_l60_used_limit_rate_min — PSI 0.0410
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.019700000062584877]11.67%16.73%5.06%0.0182
(0.019700000062584877, 0.09174999967217445]7.39%8.24%0.84%0.0009
(0.09174999967217445, 0.2787500023841858]19.24%18.81%-0.43%0.0001
(0.2787500023841858, 0.9161500036716461]49.65%41.27%-8.38%0.0155
(0.9161500036716461, inf]12.06%14.96%2.90%0.0063
apprv_amt_day — PSI 0.1320
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 11.0]79.68%63.48%-16.21%0.0369
(11.0, inf]20.32%36.52%16.21%0.0951
cre_l90_used_limit_amt_min — PSI 0.0680
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 21.640000343322754]14.76%21.47%6.71%0.0251
(21.640000343322754, 748.7099914550781]21.22%17.21%-4.01%0.0084
(748.7099914550781, 8516.22998046875]52.99%45.70%-7.30%0.0108
(8516.22998046875, 14597.384765625]5.96%6.57%0.61%0.0006
(14597.384765625, inf]5.08%9.06%3.99%0.0231
cre_now_l60_used_amt_m_min — PSI 0.0710
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 1.7799999713897705]74.09%61.70%-12.38%0.0227
(1.7799999713897705, inf]25.91%38.30%12.38%0.0484
apprv_num_m1 — PSI 0.1073
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]73.93%58.74%-15.19%0.0349
(0.5, 1.5]14.05%24.06%10.01%0.0539
(1.5, inf]12.02%17.20%5.18%0.0185
s_ps_od_days_max — PSI 0.1764
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]54.12%74.07%19.95%0.0626
(0.5, inf]45.88%25.93%-19.95%0.1138
available_limit_amt — PSI 0.0068
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.03499999921768904]32.52%32.39%-0.14%0.0000
(0.03499999921768904, 27316.265625]62.42%60.62%-1.80%0.0005
(27316.265625, inf]5.05%6.99%1.94%0.0063
cre_now_l30_used_amt_d_min — PSI 0.0345
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 1.0000500082969666]86.91%80.18%-6.73%0.0054
(1.0000500082969666, 1.382200002670288]5.38%7.35%1.96%0.0061
(1.382200002670288, inf]7.71%12.47%4.76%0.0229
max_od_days — PSI 0.2317
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 5.5]82.44%96.20%13.76%0.0212
(5.5, inf]17.56%3.80%-13.76%0.2105
setl_prcp_ratio — PSI 0.2377
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.035099999979138374]21.86%44.47%22.61%0.1605
(0.035099999979138374, inf]78.14%55.53%-22.61%0.0772
end_loan_to_cur_freq — PSI 0.1153
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, -22.5]78.67%63.41%-15.26%0.0329
(-22.5, inf]21.33%36.59%15.26%0.0824
loan_os_prcp_credit_amount_m12 — PSI 0.0786
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 16.18000030517578]8.31%12.94%4.63%0.0205
(16.18000030517578, 44.21000099182129]15.85%21.83%5.98%0.0192
(44.21000099182129, 77.875]37.97%32.44%-5.53%0.0087
(77.875, 95.76499938964844]17.43%11.02%-6.41%0.0294
(95.76499938964844, inf]20.43%21.76%1.33%0.0008
cre_now_l60_used_amt_d_min — PSI 0.0585
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 1.000950038433075]81.74%71.70%-10.04%0.0132
(1.000950038433075, 1.4549500346183777]6.89%9.65%2.76%0.0093
(1.4549500346183777, inf]11.37%18.65%7.28%0.0360
cre_now_l30_limit_left_d_max — PSI 0.0105
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.9999499917030334]19.08%23.25%4.17%0.0083
(0.9999499917030334, inf]80.92%76.75%-4.17%0.0022
shd_pre_repay_due_rate — PSI 0.2464
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.868800014257431]27.81%9.26%-18.55%0.2040
(0.868800014257431, inf]72.19%90.74%18.55%0.0424
loan_os_prcp_gt0_m3_num_pc — PSI 0.0255
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 68.16500091552734]8.81%13.83%5.02%0.0226
(68.16500091552734, inf]91.19%86.17%-5.02%0.0028
loan_os_prcp_credit_amount_m6 — PSI 0.0356
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 33.795000076293945]21.35%29.55%8.19%0.0266
(33.795000076293945, inf]78.65%70.45%-8.19%0.0090
ps_int_amt_sum — PSI 0.2234
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 209.51499938964844]8.33%24.84%16.51%0.1804
(209.51499938964844, 815.2900085449219]27.20%19.86%-7.33%0.0230
(815.2900085449219, 1891.7449951171875]27.39%20.94%-6.45%0.0173
(1891.7449951171875, 10533.93994140625]29.72%26.96%-2.75%0.0027
(10533.93994140625, inf]7.37%7.39%0.03%0.0000
useratemore50maxcontimonsm12 — PSI 0.1876
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]10.71%20.26%9.55%0.0609
(0.5, 2.5]11.27%21.66%10.39%0.0678
(2.5, inf]78.02%58.08%-19.94%0.0588
avg_tnr — PSI 0.0410
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 3.5]6.40%11.49%5.10%0.0298
(3.5, 5.5]9.64%9.21%-0.43%0.0002
(5.5, 6.5]6.99%8.13%1.15%0.0017
(6.5, 10.5]33.27%27.97%-5.30%0.0092
(10.5, inf]43.70%43.19%-0.51%0.0001
cre_now_l60_used_amt_m_max — PSI 0.2386
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 83.30500030517578]13.52%32.92%19.40%0.1726
(83.30500030517578, 2003.010009765625]63.28%44.54%-18.75%0.0659
(2003.010009765625, inf]23.19%22.54%-0.65%0.0002
max_loan_os_prcp_m12 — PSI 0.0637
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 1747.2849731445312]10.78%16.33%5.56%0.0231
(1747.2849731445312, 13572.5849609375]66.44%54.58%-11.86%0.0233
(13572.5849609375, 20848.58984375]9.51%10.92%1.41%0.0020
(20848.58984375, inf]13.28%18.17%4.89%0.0154
setl_rem_prcp_paym_m3_num — PSI 0.0163
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]84.32%79.42%-4.90%0.0029
(0.5, inf]15.68%20.58%4.90%0.0133
orig_prcp_avg_m12 — PSI 0.0373
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 1031.9722290039062]30.19%28.38%-1.82%0.0011
(1031.9722290039062, 1871.875]18.19%14.83%-3.35%0.0068
(1871.875, 3819.857177734375]22.92%21.14%-1.78%0.0014
(3819.857177734375, 13787.5]23.40%26.11%2.71%0.0030
(13787.5, inf]5.31%9.55%4.24%0.0249
useratemore70_m6_num — PSI 0.0784
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]28.42%37.70%9.28%0.0262
(0.5, 1.5]11.65%15.06%3.41%0.0088
(1.5, 4.5]37.87%33.31%-4.55%0.0058
(4.5, 5.5]9.11%5.58%-3.53%0.0173
(5.5, inf]12.95%8.34%-4.61%0.0203
useratemore70monthsubcurmonth — PSI 0.0847
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 4.5]66.81%59.30%-7.51%0.0090
(4.5, 9.5]15.86%11.51%-4.35%0.0139
(9.5, inf]17.33%29.18%11.86%0.0618
suc_apprv_num_m6 — PSI 0.0130
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 1.5]57.80%52.24%-5.56%0.0056
(1.5, 3.5]20.51%24.00%3.49%0.0055
(3.5, inf]21.68%23.76%2.07%0.0019
shd_pre_repay_amt_rate — PSI 0.0366
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.6238499879837036]33.20%41.50%8.30%0.0185
(0.6238499879837036, 0.7506499886512756]22.68%16.98%-5.70%0.0165
(0.7506499886512756, inf]44.12%41.52%-2.60%0.0016
useratemore90maxcontimonsm12 — PSI 0.1030
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]24.43%37.30%12.87%0.0544
(0.5, 1.5]18.14%19.61%1.47%0.0012
(1.5, 4.5]38.75%31.06%-7.70%0.0170
(4.5, 7.5]12.22%7.37%-4.85%0.0246
(7.5, inf]6.46%4.67%-1.79%0.0058
ps_int_amt_avg — PSI 0.1191
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 4.333299875259399]10.72%22.49%11.76%0.0871
(4.333299875259399, 13.080849647521973]32.11%24.75%-7.36%0.0192
(13.080849647521973, 38.24515151977539]37.69%31.47%-6.21%0.0112
(38.24515151977539, inf]19.48%21.30%1.81%0.0016
repay_amount_m6_all_shd_amt_pc — PSI 0.0382
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 93.98500061035156]34.83%41.73%6.90%0.0125
(93.98500061035156, 99.48500061035156]51.87%42.14%-9.74%0.0202
(99.48500061035156, inf]13.29%16.13%2.83%0.0055
cre_l30_l90_used_amt_m_min — PSI 0.0504
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.6299999952316284]78.43%68.57%-9.86%0.0132
(0.6299999952316284, inf]21.57%31.43%9.86%0.0371
cre_now_p1m_used_rate_m — PSI 0.0095
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, -4.999999873689376e-05]19.31%23.29%3.98%0.0075
(-4.999999873689376e-05, inf]80.69%76.71%-3.98%0.0020
useratemore90_m3_num — PSI 0.0027
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]61.13%59.31%-1.82%0.0006
(0.5, 1.5]16.07%17.37%1.29%0.0010
(1.5, 2.5]10.45%11.47%1.01%0.0009
(2.5, inf]12.34%11.86%-0.49%0.0002
min_apply_amt — PSI 0.0266
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 1135.0]8.90%12.49%3.59%0.0122
(1135.0, 4180.0]47.22%41.64%-5.59%0.0070
(4180.0, 7300.0]24.21%22.57%-1.65%0.0012
(7300.0, inf]19.66%23.31%3.65%0.0062
cre_now_l30_limit_left_d_min — PSI 0.0026
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.997050017118454]16.82%14.97%-1.85%0.0022
(0.997050017118454, inf]83.18%85.03%1.85%0.0004
loan_os_prcp_gt0_m12_num_pc — PSI 0.0334
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 57.73500061035156]8.67%14.16%5.50%0.0270
(57.73500061035156, 72.08000183105469]6.91%7.87%0.96%0.0012
(72.08000183105469, inf]84.42%77.97%-6.45%0.0051
cre_l30_l90_usedrate_m_max — PSI 0.1840
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.025150000117719173]31.21%51.40%20.19%0.1007
(0.025150000117719173, 0.0799499973654747]8.70%8.76%0.07%0.0000
(0.0799499973654747, inf]60.10%39.84%-20.26%0.0833
cre_now_l60_limit_left_d_min — PSI 0.0108
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.9717499911785126]22.39%18.22%-4.17%0.0086
(0.9717499911785126, inf]77.61%81.78%4.17%0.0022
min_tnr — PSI 0.0021
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 5.0]60.27%57.99%-2.28%0.0009
(5.0, inf]39.73%42.01%2.28%0.0013
last_clear_num_m1 — PSI 0.0046
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, 0.5]79.20%81.89%2.69%0.0009
(0.5, 1.5]12.91%11.22%-1.69%0.0024
(1.5, inf]7.89%6.88%-1.00%0.0014
setl_rem_prcp_paym_last_month_lo — PSI 0.1082
分栏开发基准比例PSI时点样本 比例变更贡献
(-inf, -4999.0]7.62%18.39%10.77%0.0949
(-4999.0, inf]92.38%81.61%-10.77%0.0133

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。

due_day_my
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 30.5]3493791.73%3493700.00%—
2(30.5, inf]3151100.00%03151100.00%0.00
合计380883493731518.27%11.09
2026-10-27T09:47:32.433083 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(0.00% → 100.00%),WOE 单调下降(8.7486 → -11.1544),符合风控分箱预期。IV=19.9031,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((30.5, inf])为极高风险区间:坏率高达 100.00%,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]2042253.63%18144227811.15%7.96
2(0.5, inf]17658100.00%167868724.94%19.25
合计380803493031508.27%11.09
2026-10-27T09:47:32.536071 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(11.15% → 4.94%),WOE 单调上升(-0.3309 → 0.5516),符合风控分箱预期。IV=0.1798,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 1 箱((-inf, 0.5])为坏率最高区间:坏率高达 11.15%,可作为拒绝或重点风控策略的阈值参考

suc_apprv_num_m1
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]3016979.23%2728728829.55%9.47
2(0.5, 1.5]478791.80%45911964.09%23.42
3(1.5, inf]3124100.00%3052722.30%42.39
合计380803493031508.27%11.09
2026-10-27T09:47:32.710646 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(9.55% → 2.30%),WOE 单调上升(-0.1580 → 1.3409),符合风控分箱预期。IV=0.1594,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])为坏率最高区间:坏率高达 9.55%,可作为拒绝或重点风控策略的阈值参考

cre_l60_used_limit_rate_min
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.019700000062584877]444111.67%43061353.04%31.90
2(0.019700000062584877, 0.09174999967217445]281319.06%26561575.58%16.92
3(0.09174999967217445, 0.2787500023841858]732338.30%68075167.05%13.19
4(0.2787500023841858, 0.9161500036716461]1889987.94%1715717429.22%9.85
5(0.9161500036716461, inf]4589100.00%398960013.07%6.65
合计380653491531508.28%11.08
2026-10-27T09:47:32.834864 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(3.04% → 13.07%),WOE 单调下降(1.0570 → -0.5111),符合风控分箱预期。IV=0.1478,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 5 箱((0.9161500036716461, inf])为坏率最高区间:坏率高达 13.07%,可作为拒绝或重点风控策略的阈值参考

apprv_amt_day
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 11.0]3034479.68%2745228929.53%9.49
2(11.0, inf]7736100.00%74782583.34%28.98
合计380803493031508.27%11.09
2026-10-27T09:47:32.953557 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(9.53% → 3.34%),WOE 单调上升(-0.1555 → 0.9608),符合风控分箱预期。IV=0.1475,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 1 箱((-inf, 11.0])为坏率最高区间:坏率高达 9.53%,可作为拒绝或重点风控策略的阈值参考

cre_l90_used_limit_amt_min
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 21.640000343322754]561814.76%54082103.74%25.75
2(21.640000343322754, 748.7099914550781]807635.98%75265506.81%13.68
3(748.7099914550781, 8516.22998046875]2017288.97%1835418189.01%10.10
4(8516.22998046875, 14597.384765625]226794.92%199027712.22%7.18
5(14597.384765625, inf]1932100.00%163729515.27%5.55
合计380653491531508.28%11.08
2026-10-27T09:47:33.071590 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(3.74% → 15.27%),WOE 单调下降(0.8430 → -0.6919),符合风控分箱预期。IV=0.1336,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 5 箱((14597.384765625, inf])为坏率最高区间:坏率高达 15.27%,可作为拒绝或重点风控策略的阈值参考

cre_now_l60_used_amt_m_min
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 1.7799999713897705]2820174.09%2545827439.73%9.28
2(1.7799999713897705, inf]9864100.00%94574074.13%23.24
合计380653491531508.28%11.08
2026-10-27T09:47:33.188862 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(9.73% → 4.13%),WOE 单调上升(-0.1775 → 0.7402),符合风控分箱预期。IV=0.1300,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 1 箱((-inf, 1.7799999713897705])为坏率最高区间:坏率高达 9.73%,可作为拒绝或重点风控策略的阈值参考

apprv_num_m1
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]2815273.93%2542527279.69%9.32
2(0.5, 1.5]535087.98%51072434.54%21.02
3(1.5, inf]4578100.00%43981803.93%24.43
合计380803493031508.27%11.09
2026-10-27T09:47:33.295077 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(9.69% → 3.93%),WOE 单调上升(-0.1734 → 0.7900),符合风控分箱预期。IV=0.1224,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])为坏率最高区间:坏率高达 9.69%,可作为拒绝或重点风控策略的阈值参考

s_ps_od_days_max
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]2061054.12%18430218010.58%8.45
2(0.5, inf]17470100.00%165009705.55%17.01
合计380803493031508.27%11.09
2026-10-27T09:47:33.401173 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(10.58% → 5.55%),WOE 单调上升(-0.2713 → 0.4279),符合风控分箱预期。IV=0.1150,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 1 箱((-inf, 0.5])为坏率最高区间:坏率高达 10.58%,可作为拒绝或重点风控策略的阈值参考

available_limit_amt
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.03499999921768904]1238532.52%10918146711.85%7.44
2(0.03499999921768904, 27316.265625]2377194.95%2218315886.68%13.97
3(27316.265625, inf]1924100.00%1829954.94%19.25
合计380803493031508.27%11.09
2026-10-27T09:47:33.509898 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(11.85% → 4.94%),WOE 单调上升(-0.3987 → 0.5517),符合风控分箱预期。IV=0.1036,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, 0.03499999921768904])为坏率最高区间:坏率高达 11.85%,可作为拒绝或重点风控策略的阈值参考

cre_now_l30_used_amt_d_min
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 1.0000500082969666]3121186.91%2827329389.41%9.62
2(1.0000500082969666, 1.382200002670288]193392.29%1860733.78%25.48
3(1.382200002670288, inf]2768100.00%2687812.93%33.17
合计359123282030928.61%10.61
2026-10-27T09:47:33.624682 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(9.41% → 2.93%),WOE 单调上升(-0.0980 → 1.1395),符合风控分箱预期。IV=0.1011,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, 1.0000500082969666])为坏率最高区间:坏率高达 9.41%,可作为拒绝或重点风控策略的阈值参考

max_od_days
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 5.5]3139582.44%2916222337.11%13.06
2(5.5, inf]6685100.00%576891713.72%6.29
合计380803493031508.27%11.09
2026-10-27T09:47:33.742428 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(7.11% → 13.72%),WOE 单调下降(0.1636 → -0.5670),符合风控分箱预期。IV=0.0920,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 2 箱((5.5, inf])为坏率最高区间:坏率高达 13.72%,可作为拒绝或重点风控策略的阈值参考

setl_prcp_ratio
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.035099999979138374]832521.86%79603654.38%21.81
2(0.035099999979138374, inf]29755100.00%2697027859.36%9.68
合计380803493031508.27%11.09
2026-10-27T09:47:33.853771 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(4.38% → 9.36%),WOE 单调下降(0.6763 → -0.1355),符合风控分箱预期。IV=0.0909,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 2 箱((0.035099999979138374, inf])为坏率最高区间:坏率高达 9.36%,可作为拒绝或重点风控策略的阈值参考

end_loan_to_cur_freq
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, -22.5]2995978.67%2716227979.34%9.71
2(-22.5, inf]8121100.00%77683534.35%22.01
合计380803493031508.27%11.09
2026-10-27T09:47:33.964676 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(9.34% → 4.35%),WOE 单调上升(-0.1327 → 0.6854),符合风控分箱预期。IV=0.0902,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, -22.5])为坏率最高区间:坏率高达 9.34%,可作为拒绝或重点风控策略的阈值参考

loan_os_prcp_credit_amount_m12
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 16.18000030517578]31648.31%30581063.35%28.85
2(16.18000030517578, 44.21000099182129]603324.16%56713626.00%15.67
3(44.21000099182129, 77.875]1445362.14%1322312308.51%10.75
4(77.875, 95.76499938964844]663579.57%60086279.45%9.58
5(95.76499938964844, inf]7775100.00%695182410.60%8.44
合计380603491131498.27%11.09
2026-10-27T09:47:34.161716 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(3.35% → 10.60%),WOE 单调下降(0.9564 → -0.2732),符合风控分箱预期。IV=0.0894,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((95.76499938964844, inf])为坏率最高区间:坏率高达 10.60%,可作为拒绝或重点风控策略的阈值参考

cre_now_l60_used_amt_d_min
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 1.000950038433075]2794381.74%2520527389.80%9.21
2(1.000950038433075, 1.4549500346183777]235688.63%22351215.14%18.47
3(1.4549500346183777, inf]3886100.00%37151714.40%21.73
合计341853115530308.86%10.28
2026-10-27T09:47:34.280726 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(9.80% → 4.40%),WOE 单调上升(-0.1106 → 0.7481),符合风控分箱预期。IV=0.0761,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 1 箱((-inf, 1.000950038433075])为坏率最高区间:坏率高达 9.80%,可作为拒绝或重点风控策略的阈值参考

cre_now_l30_limit_left_d_max
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.9999499917030334]721419.08%68873274.53%21.06
2(0.9999499917030334, inf]30600100.00%2779828029.16%9.92
合计378143468531298.27%11.09
2026-10-27T09:47:34.389395 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(4.53% → 9.16%),WOE 单调下降(0.6418 → -0.1110),符合风控分箱预期。IV=0.0708,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 2 箱((0.9999499917030334, inf])为坏率最高区间:坏率高达 9.16%,可作为拒绝或重点风控策略的阈值参考

shd_pre_repay_due_rate
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.868800014257431]1047327.81%99205535.28%17.94
2(0.868800014257431, inf]27191100.00%2461825739.46%9.57
合计376643453831268.30%11.05
2026-10-27T09:47:34.493466 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(5.28% → 9.46%),WOE 单调下降(0.4846 → -0.1439),符合风控分箱预期。IV=0.0693,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 2 箱((0.868800014257431, inf])为坏率最高区间:坏率高达 9.46%,可作为拒绝或重点风控策略的阈值参考

loan_os_prcp_gt0_m3_num_pc
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 68.16500091552734]33548.81%32521023.04%31.88
2(68.16500091552734, inf]34706100.00%3165930478.78%10.39
合计380603491131498.27%11.09
2026-10-27T09:47:34.598481 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(3.04% → 8.78%),WOE 单调下降(1.0563 → -0.0649),符合风控分箱预期。IV=0.0681,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 2 箱((68.16500091552734, inf])为坏率最高区间:坏率高达 8.78%,可作为拒绝或重点风控策略的阈值参考

loan_os_prcp_credit_amount_m6
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 33.795000076293945]812721.35%77343934.84%19.68
2(33.795000076293945, inf]29933100.00%2717727569.21%9.86
合计380603491131498.27%11.09
2026-10-27T09:47:34.714525 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(4.84% → 9.21%),WOE 单调下降(0.5739 → -0.1171),符合风控分箱预期。IV=0.0668,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 2 箱((33.795000076293945, inf])为坏率最高区间:坏率高达 9.21%,可作为拒绝或重点风控策略的阈值参考

ps_int_amt_sum
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 209.51499938964844]31738.33%30021715.39%17.56
2(209.51499938964844, 815.2900085449219]1035635.53%96177397.14%13.01
3(815.2900085449219, 1891.7449951171875]1042962.91%96138167.82%11.78
4(1891.7449951171875, 10533.93994140625]1131692.63%1027610409.19%9.88
5(10533.93994140625, inf]2806100.00%242238413.69%6.31
合计380803493031508.27%11.09
2026-10-27T09:47:34.830052 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(5.39% → 13.69%),WOE 单调下降(0.4594 → -0.5642),符合风控分箱预期。IV=0.0558,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 5 箱((10533.93994140625, inf])为坏率最高区间:坏率高达 13.69%,可作为拒绝或重点风控策略的阈值参考

useratemore50maxcontimonsm12
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]407610.71%38891874.59%20.80
2(0.5, 2.5]428921.98%40352545.92%15.89
3(2.5, inf]29695100.00%2698727089.12%9.97
合计380603491131498.27%11.09
2026-10-27T09:47:34.949858 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(4.59% → 9.12%),WOE 单调下降(0.6291 → -0.1066),符合风控分箱预期。IV=0.0545,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 1 箱 WOE<0(风险高于平均)

第 3 箱((2.5, inf])为坏率最高区间:坏率高达 9.12%,可作为拒绝或重点风控策略的阈值参考

avg_tnr
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 3.5]24366.40%2340963.94%24.38
2(3.5, 5.5]367216.04%34422306.26%14.97
3(5.5, 6.5]266023.03%24721887.07%13.15
4(6.5, 10.5]1267156.30%1164510268.10%11.35
5(10.5, inf]16641100.00%1503116109.67%9.34
合计380803493031508.27%11.09
2026-10-27T09:47:35.067696 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(3.94% → 9.67%),WOE 单调下降(0.7876 → -0.1721),符合风控分箱预期。IV=0.0524,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 5 箱((10.5, inf])为坏率最高区间:坏率高达 9.67%,可作为拒绝或重点风控策略的阈值参考

cre_now_l60_used_amt_m_max
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 83.30500030517578]514813.52%49042444.74%20.10
2(83.30500030517578, 2003.010009765625]2408876.81%2208320058.32%11.01
3(2003.010009765625, inf]8829100.00%792890110.21%8.80
合计380653491531508.28%11.08
2026-10-27T09:47:35.181389 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(4.74% → 10.21%),WOE 单调下降(0.5951 → -0.2309),符合风控分箱预期。IV=0.0511,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((2003.010009765625, inf])为坏率最高区间:坏率高达 10.21%,可作为拒绝或重点风控策略的阈值参考

max_loan_os_prcp_m12
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 1747.2849731445312]410110.78%38602415.88%16.02
2(1747.2849731445312, 13572.5849609375]2528677.21%2332819587.74%11.91
3(13572.5849609375, 20848.58984375]361986.72%32783419.42%9.61
4(20848.58984375, inf]5054100.00%444560912.05%7.30
合计380603491131498.27%11.09
2026-10-27T09:47:35.302536 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(5.88% → 12.05%),WOE 单调下降(0.3679 → -0.4180),符合风控分箱预期。IV=0.0455,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 2 箱 WOE<0(风险高于平均)

第 4 箱((20848.58984375, inf])为坏率最高区间:坏率高达 12.05%,可作为拒绝或重点风控策略的阈值参考

setl_rem_prcp_paym_m3_num
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]3209484.32%2924028548.89%10.25
2(0.5, inf]5966100.00%56712954.94%19.22
合计380603491131498.27%11.09
2026-10-27T09:47:35.489192 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(8.89% → 4.94%),WOE 单调上升(-0.0789 → 0.5504),符合风控分箱预期。IV=0.0433,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])为坏率最高区间:坏率高达 8.89%,可作为拒绝或重点风控策略的阈值参考

orig_prcp_avg_m12
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 1031.9722290039062]1149830.19%107357636.64%14.07
2(1031.9722290039062, 1871.875]692548.38%64055207.51%12.32
3(1871.875, 3819.857177734375]872771.30%79937348.41%10.89
4(3819.857177734375, 13787.5]890994.69%80418689.74%9.26
5(13787.5, inf]2021100.00%175626513.11%6.63
合计380803493031508.27%11.09
2026-10-27T09:47:35.613368 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(6.64% → 13.11%),WOE 单调下降(0.2381 → -0.5149),符合风控分箱预期。IV=0.0431,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((13787.5, inf])为坏率最高区间:坏率高达 13.11%,可作为拒绝或重点风控策略的阈值参考

useratemore70_m6_num
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]1081728.42%101466716.20%15.12
2(0.5, 1.5]443540.07%41153207.22%12.86
3(1.5, 4.5]1441277.94%1310413089.08%10.02
4(4.5, 5.5]346887.05%31323369.69%9.32
5(5.5, inf]4928100.00%441451410.43%8.59
合计380603491131498.27%11.09
2026-10-27T09:47:35.739541 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(6.20% → 10.43%),WOE 单调下降(0.3103 → -0.2554),符合风控分箱预期。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 5 箱((5.5, inf])为坏率最高区间:坏率高达 10.43%,可作为拒绝或重点风控策略的阈值参考

useratemore70monthsubcurmonth
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 4.5]2542966.81%2310223279.15%9.93
2(4.5, 9.5]603782.67%55644737.83%11.76
3(9.5, inf]6594100.00%62453495.29%17.89
合计380603491131498.27%11.09
2026-10-27T09:47:35.859489 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(9.15% → 5.29%),WOE 单调上升(-0.1104 → 0.4787),符合风控分箱预期。IV=0.0416,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 1 箱((-inf, 4.5])为坏率最高区间:坏率高达 9.15%,可作为拒绝或重点风控策略的阈值参考

suc_apprv_num_m6
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 1.5]2201157.80%1992520869.48%9.55
2(1.5, 3.5]781278.32%72405727.32%12.66
3(3.5, inf]8257100.00%77654925.96%15.78
合计380803493031508.27%11.09
2026-10-27T09:47:35.975248 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(9.48% → 5.96%),WOE 单调上升(-0.1492 → 0.3530),符合风控分箱预期。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 2 箱 WOE<0(风险高于平均)

第 1 箱((-inf, 1.5])为坏率最高区间:坏率高达 9.48%,可作为拒绝或重点风控策略的阈值参考

shd_pre_repay_amt_rate
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.6238499879837036]1263733.20%118148236.51%14.35
2(0.6238499879837036, 0.7506499886512756]863155.88%79446877.96%11.56
3(0.7506499886512756, inf]16791100.00%1515216399.76%9.24
合计380593491031498.27%11.09
2026-10-27T09:47:36.088141 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(6.51% → 9.76%),WOE 单调下降(0.2584 → -0.1816),符合风控分箱预期。IV=0.0360,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 1 箱 WOE<0(风险高于平均)

第 3 箱((0.7506499886512756, inf])为坏率最高区间:坏率高达 9.76%,可作为拒绝或重点风控策略的阈值参考

useratemore90maxcontimonsm12
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]929824.43%87235756.18%15.17
2(0.5, 1.5]690342.57%63495548.03%11.46
3(1.5, 4.5]1475081.32%1346112898.74%10.44
4(4.5, 7.5]465293.54%41884649.97%9.03
5(7.5, inf]2457100.00%219026710.87%8.20
合计380603491131498.27%11.09
2026-10-27T09:47:36.209216 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(6.18% → 10.87%),WOE 单调下降(0.3136 → -0.3013),符合风控分箱预期。IV=0.0350,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((7.5, inf])为坏率最高区间:坏率高达 10.87%,可作为拒绝或重点风控策略的阈值参考

ps_int_amt_avg
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 4.333299875259399]408310.72%38572265.54%17.07
2(4.333299875259399, 13.080849647521973]1222742.83%113149137.47%12.39
3(13.080849647521973, 38.24515151977539]1435180.52%1310612458.68%10.53
4(38.24515151977539, inf]7419100.00%665376610.32%8.69
合计380803493031508.27%11.09
2026-10-27T09:47:36.329610 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(5.54% → 10.32%),WOE 单调下降(0.4312 → -0.2443),符合风控分箱预期。IV=0.0344,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 2 箱 WOE<0(风险高于平均)

第 4 箱((38.24515151977539, inf])为坏率最高区间:坏率高达 10.32%,可作为拒绝或重点风控策略的阈值参考

repay_amount_m6_all_shd_amt_pc
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 93.98500061035156]1325734.83%1194613119.89%9.11
2(93.98500061035156, 99.48500061035156]1974386.71%1819315507.85%11.74
3(99.48500061035156, inf]5060100.00%47722885.69%16.57
合计380603491131498.27%11.09
2026-10-27T09:47:36.444331 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(9.89% → 5.69%),WOE 单调上升(-0.1961 → 0.4018),符合风控分箱预期。IV=0.0344,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 1 箱((-inf, 93.98500061035156])为坏率最高区间:坏率高达 9.89%,可作为拒绝或重点风控策略的阈值参考

cre_l30_l90_used_amt_m_min
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.6299999952316284]2985678.43%2718526718.95%10.18
2(0.6299999952316284, inf]8209100.00%77304795.84%16.14
合计380653491531508.28%11.08
2026-10-27T09:47:36.552733 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(8.95% → 5.84%),WOE 单调上升(-0.0853 → 0.3756),符合风控分箱预期。IV=0.0320,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, 0.6299999952316284])为坏率最高区间:坏率高达 8.95%,可作为拒绝或重点风控策略的阈值参考

cre_now_p1m_used_rate_m
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, -4.999999873689376e-05]728819.31%68704185.74%16.44
2(-4.999999873689376e-05, inf]30446100.00%2773527118.90%10.23
合计377343460531298.29%11.06
2026-10-27T09:47:36.661177 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(5.74% → 8.90%),WOE 单调下降(0.3962 → -0.0779),符合风控分箱预期。IV=0.0308,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 2 箱((-4.999999873689376e-05, inf])为坏率最高区间:坏率高达 8.90%,可作为拒绝或重点风控策略的阈值参考

useratemore90_m3_num
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]2326661.13%2150117657.59%12.18
2(0.5, 1.5]611777.20%56464717.70%11.99
3(1.5, 2.5]397987.66%36233568.95%10.18
4(2.5, inf]4698100.00%414155711.86%7.43
合计380603491131498.27%11.09
2026-10-27T09:47:36.773447 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(7.59% → 11.86%),WOE 单调下降(0.0942 → -0.3996),符合风控分箱预期。IV=0.0302,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 2 箱 WOE<0(风险高于平均)

第 4 箱((2.5, inf])为坏率最高区间:坏率高达 11.86%,可作为拒绝或重点风控策略的阈值参考

min_apply_amt
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 1135.0]33888.90%301037811.16%7.96
2(1135.0, 4180.0]1798356.12%1641315708.73%10.45
3(4180.0, 7300.0]922180.34%84907317.93%11.61
4(7300.0, inf]7488100.00%70174716.29%14.90
合计380803493031508.27%11.09
2026-10-27T09:47:36.955858 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(11.16% → 6.29%),WOE 单调上升(-0.3311 → 0.2953),符合风控分箱预期。IV=0.0286,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 2 箱 WOE<0(风险高于平均)

第 1 箱((-inf, 1135.0])为坏率最高区间:坏率高达 11.16%,可作为拒绝或重点风控策略的阈值参考

cre_now_l30_limit_left_d_min
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.997050017118454]613616.82%544369311.29%7.85
2(0.997050017118454, inf]30339100.00%2797523647.79%11.83
合计364753341830578.38%10.93
2026-10-27T09:47:37.064259 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(11.29% → 7.79%),WOE 单调上升(-0.3306 → 0.0793),符合风控分箱预期。IV=0.0262,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, 0.997050017118454])为坏率最高区间:坏率高达 11.29%,可作为拒绝或重点风控策略的阈值参考

loan_os_prcp_gt0_m12_num_pc
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 57.73500061035156]32988.67%31381604.85%19.61
2(57.73500061035156, 72.08000183105469]263115.58%24172148.13%11.29
3(72.08000183105469, inf]32131100.00%2935627758.64%10.58
合计380603491131498.27%11.09
2026-10-27T09:47:37.178104 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(4.85% → 8.64%),WOE 单调下降(0.5705 → -0.0469),符合风控分箱预期。IV=0.0242,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 1 箱 WOE<0(风险高于平均)

第 3 箱((72.08000183105469, inf])为坏率最高区间:坏率高达 8.64%,可作为拒绝或重点风控策略的阈值参考

cre_l30_l90_usedrate_m_max
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.025150000117719173]1187931.21%110807996.73%13.87
2(0.025150000117719173, 0.0799499973654747]331039.90%30532577.76%11.88
3(0.0799499973654747, inf]22876100.00%2078220949.15%9.92
合计380653491531508.28%11.08
2026-10-27T09:47:37.294682 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(6.73% → 9.15%),WOE 单调下降(0.2240 → -0.1105),符合风控分箱预期。IV=0.0224,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 1 箱 WOE<0(风险高于平均)

第 3 箱((0.0799499973654747, inf])为坏率最高区间:坏率高达 9.15%,可作为拒绝或重点风控策略的阈值参考

cre_now_l60_limit_left_d_min
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.9717499911785126]803922.39%718685310.61%8.42
2(0.9717499911785126, inf]27862100.00%2569321697.78%11.85
合计359013287930228.42%10.88
2026-10-27T09:47:37.407628 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(10.61% → 7.78%),WOE 单调上升(-0.2558 → 0.0850),符合风控分箱预期。IV=0.0217,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, 0.9717499911785126])为坏率最高区间:坏率高达 10.61%,可作为拒绝或重点风控策略的阈值参考

min_tnr
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 5.0]2294960.27%2126016897.36%12.59
2(5.0, inf]15131100.00%1367014619.66%9.36
合计380803493031508.27%11.09
2026-10-27T09:47:37.512924 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(7.36% → 9.66%),WOE 单调下降(0.1267 → -0.1699),符合风控分箱预期。IV=0.0215,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 2 箱((5.0, inf])为坏率最高区间:坏率高达 9.66%,可作为拒绝或重点风控策略的阈值参考

last_clear_num_m1
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, 0.5]3016179.20%2750926528.79%10.37
2(0.5, 1.5]491692.11%45943226.55%14.27
3(1.5, inf]3003100.00%28271765.86%16.06
合计380803493031508.27%11.09
2026-10-27T09:47:37.615436 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(8.79% → 5.86%),WOE 单调上升(-0.0667 → 0.3705),符合风控分箱预期。IV=0.0203,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 1 箱((-inf, 0.5])为坏率最高区间:坏率高达 8.79%,可作为拒绝或重点风控策略的阈值参考

setl_rem_prcp_paym_last_month_lo
组号ValueTot. Records累计占比#Goods#BadsBad RateG/B Odds
1(-inf, -4999.0]28997.62%27571424.90%19.42
2(-4999.0, inf]35161100.00%3215430078.55%10.69
合计380603491131498.27%11.09
2026-10-27T09:47:37.723758 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(4.90% → 8.55%),WOE 单调下降(0.5604 → -0.0361),符合风控分箱预期。IV=0.0202,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 2 箱((-4999.0, inf])为坏率最高区间:坏率高达 8.55%,可作为拒绝或重点风控策略的阈值参考

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(跨时间样本)与情景2(跨时间样本剔除一半好样本,批注16)。

测试情景1:跨时间验证样本

变量贡献度

变量贡献度
due_day_my7.3878
suc_apprv_num_m10.0065
f_ps_od_days_max0.0061
cre_l60_used_limit_rate_min0.0051
cre_l90_used_limit_amt_min0.0040

区分度验证 — LR + WOE 评分卡 @ 跨时间样本

KS = 1.0000,GINI = 1.0000

2026-10-27T09:47:37.865422 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-10-27T09:47:37.989709 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

测试情景2:跨时间样本剔除一半好样本

指标数值
样本量34730 → 18581(剔除好样本 16149)
KS1.0000
AUC1.0000
GINI1.0000
2026-10-27T09:47:38.114146 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/ 2026-10-27T09:47:38.306001 image/svg+xml Matplotlib v3.10.9, https://matplotlib.org/

注:情景2 模拟好样本分布右移的压力情形(固定 seed=42,可复现);剔除后 KS/AUC 下降幅度反映模型对好样本结构的敏感度。