评分卡建模自动化流程
- 本项目分为2个脚本,一个是评分卡建模主流程(评分卡模板GitHub_v1.ipynb),一个是主流程里需要用到的一些函数(self_f_2.py);此脚本**评分卡模版_github待完善.py **待更新完善
- 主流程涉及的代码流程为从数据基本情况了解(简单eda)——数据预处理(异常和缺失)——特征衍生——特征筛选——模型评价——分数映射——分数分布,关键步骤的数据都进行了输出展示,以让建模者能够根据自身数据情况,进行调整
- 主流程里需要用到的一些函数,基本参考了GitHub项目https://github.com/taenggu0309/Semi-auto-modeling 笔者做了些改造和新增;
- ipynb(使用浏览器打开)文件可能不好打开,可以下载到本地打开查看
- 创建评分卡建模自动化流程项目
- 进行细微修正
- 计算变量得分时,刻度B少添加一个负号,导致分数错误,现已修正;
- 由于存在需要知道具体某个用户得分的情况,在最后添加里将用户唯一识别码合并到数据集里到操作
- 由于部分网友,想要训练的数据集,因此添加了数据集
- 进行代码添加
- 细化调整分箱部分的代码,添加手动调整分箱的代码
- 去除在woe编码前的共线性筛选,以免去除掉一些因为woe编码后因为相关性筛选而去除的变量