风控模型实验报告

实验 ID: 20260712-154652-d7c1 · 生成时间: 2026-07-12 07:46 UTC

1. 数据审计

样本数20
字段数9
坏样本率40.00%
高风险字段2

审计建议

  • 列 `\f0\fs24 \cf0 user_id` 唯一率过高(1.0)，可能是 ID 列，应排除在特征之外
  • 列 `age` 唯一率过高(1.0)，可能是 ID 列，应排除在特征之外

2. 样本拆分

拆分方式: 时间切分，总样本 20，目标非空 20

数据集样本数比例坏样本率时间范围
训练集1260.00%41.67%—
验证集420.00%50.00%—
OOT 集420.00%25.00%—

3. 变量筛选

候选变量 7，入模 6，IV 阈值 0.01

变量IV单调分箱数
overdue_count1.6197否4
income1.1786否4
job_years1.1050否3
debt_ratio0.7648否10
credit_score0.6931是10
loan_amount0.4249否4

4. 模型评估

模型数据集AUCKSLift@10%PSI
LR + WOE 评分卡训练集1.00001.00002.40—
LR + WOE 评分卡验证集1.00001.00002.00—
LR + WOE 评分卡OOT 集1.00001.00004.007.4236

LR + WOE 评分卡 — 训练集 坏样本率分档

分组样本数坏样本坏样本率分数范围
1200.00%0.08 ~ 0.08
2100.00%0.08 ~ 0.08
4200.00%0.09 ~ 0.09
6200.00%0.09 ~ 0.09
711100.00%0.68 ~ 0.68
811100.00%0.82 ~ 0.82
911100.00%0.87 ~ 0.87
1022100.00%0.88 ~ 0.88

LR + WOE 评分卡 — 验证集 坏样本率分档

分组样本数坏样本坏样本率分数范围
1100.00%0.09 ~ 0.09
4100.00%0.20 ~ 0.20
711100.00%0.88 ~ 0.88
1011100.00%0.91 ~ 0.91

LR + WOE 评分卡 — OOT 集 坏样本率分档

分组样本数坏样本坏样本率分数范围
1100.00%0.08 ~ 0.08
7200.00%0.12 ~ 0.12
1011100.00%0.86 ~ 0.86

5. 评分卡

变量分箱WOE系数分数
income(-inf, 3450.0]-0.4055-0.507574
income(3450.0, 4400.0]-0.4055-0.507574
income(4400.0, 5000.0]-0.5596-0.507572
income(5000.0, inf]0.8755-0.507593
job_years(-inf, 4.700000000000001]-0.6568-0.559669
job_years(4.700000000000001, 18.200000000000003]0.8183-0.559693
job_years(18.200000000000003, inf]-0.4055-0.559673
loan_amount(-inf, 45000.0]0.1054-0.423081
loan_amount(45000.0, 48500.00000000001]0.0000-0.423080
loan_amount(48500.00000000001, 60000.0]0.4700-0.423086
loan_amount(60000.0, inf]-0.5232-0.423074
credit_score(-inf, 559.0]-0.4055-0.526474
credit_score(559.0, 588.0]-0.4055-0.526474
credit_score(588.0, 607.0]-0.4055-0.526474
credit_score(607.0, 632.0]-0.4055-0.526474
credit_score(632.0, 655.0]0.2877-0.526484
credit_score(655.0, 674.0]0.2877-0.526484
credit_score(674.0, 693.0]0.2877-0.526484
credit_score(693.0, 702.0]0.2877-0.526484
credit_score(702.0, 721.0]0.2877-0.526484
credit_score(721.0, inf]0.2877-0.526484
overdue_count(-inf, 1.0]1.0986-1.1067115
overdue_count(1.0, 2.0]-0.5596-1.106762
overdue_count(2.0, 3.2857142857142847]-0.4055-1.106767
overdue_count(3.2857142857142847, inf]-0.5596-1.106762
debt_ratio(-inf, 0.295]0.2877-0.587085
debt_ratio(0.295, 0.316]0.2877-0.587085
debt_ratio(0.316, 0.35]0.4055-0.587087
debt_ratio(0.35, 0.43000000000000005]0.1542-0.587083
debt_ratio(0.43000000000000005, 0.45]0.4055-0.587087
debt_ratio(0.45, 0.5400000000000003]0.1542-0.587083
debt_ratio(0.5400000000000003, 0.65]-0.5596-0.587070
debt_ratio(0.65, 0.71]-0.2231-0.587076
debt_ratio(0.71, 0.8050000000000002]-0.4055-0.587073
debt_ratio(0.8050000000000002, inf]-0.4055-0.587073