iv计算code

import numpy as np
import pandas as pd

data={"feature1":[1,1,2,1,3,2,3],"feature2":[1,3,2,1,2,2,3],"label":[0,0,0,1,1,0,1]}
df=pd.DataFrame(data)

def calcute_iv(df,aim,res,py_i_all,pn_i_all):
    '''
    data : 只包含feature，和label
    aim ：关心的相应变量值，（流失为1）
    '''
    iv = 0
    feature,label = df.columns
    for i,data in df.groupby(df[feature]):
        py_i = (data[label] == aim).sum() + 1
        pn_i = (data[label] != aim).sum() + 1
        woe_i = np.log((py_i/py_i_all)/(pn_i/pn_i_all))
        iv += woe_i
    res[feature] = iv

def fun_iv(train_data,label,aim):
    '''
    train_data : 用于计算的基础数据，包括label。由于没有做分组，因此特征需为离散特征
    label : 指定数据集中label的名称
    aim ：关心的相应变量值，（流失为1）
    return :
        res : 各个特征计算iv结果对于的一个字典
    '''
    res = dict()
    columns = list(train_data.columns)
    py_i_all = (train_data[label]==aim).sum() + 1
    pn_i_all = (train_data[label]!=aim).sum() + 1
    columns.remove(label)    
    for column in columns:
        data = train_data[[column,label]]
        calcute_iv(data,aim,res,py_i_all,pn_i_all)
    return res
res = fun_iv(df,'label',1)
res  ---- >>>
{'feature1': -0.4054651081081644, 'feature2': -0.40546510810816444}