deburky/home-credit-credit-risk-model-stability

Home Credit - 信用风险模型稳定性（已处理）数据集是来自Kaggle竞赛的已处理子集，专门为关于WoE、IV和PSI信息理论基础的文章准备。数据集包含522,596笔贷款申请，48列（32个预测变量、目标变量、元数据和11个申请人/信用局属性）。目标变量是二元违约指标（1=违约，0=非违约）。数据来源包括`train_base`、`train_static_0_1`、`train_static_cb_0`和`train_person_1`表。数据集包含申请人属性（如性别、年龄、教育水平、收入类型等）和信用局属性（如婚姻状况、请求类型等）。特征组包括逾期天数、信用局查询和用户档案等。原始数据集包含深度=0（直接与`case_id`关联的静态特征）和深度=1（需要聚合的历史记录）的表。

The Home Credit - Credit Risk Model Stability (Processed) dataset is a processed subset from the Kaggle competition, prepared for the RWDS article on information-theoretic foundations of WoE, IV, and PSI. It contains 522,596 loan applications with 48 columns (32 predictors + target + metadata + 11 applicant/CB attributes). The target is a binary default indicator (1 = default, 0 = non-default). Source tables include `train_base`, `train_static_0_1`, `train_static_cb_0`, and `train_person_1`. The dataset includes applicant attributes (e.g., sex, age, education level, income type) and credit bureau attributes (e.g., marital status, request type). Feature groups cover Days Past Due, Credit Bureau, and User Profile. The original dataset includes depth=0 tables (static features tied directly to `case_id`) and depth=1 tables (historical records requiring aggregation).