使用scorecardpy库计算woe分箱和iv值

woe分箱_iv值计算

• 基于scorecardpy库，乳腺癌数据集

  import pandas as pd
  import numpy as np
  from sklearn.datasets import load_breast_cancer
  import scorecardpy as sc
  from tqdm import notebook
  
  cancer = load_breast_cancer()
  df = pd.DataFrame(cancer.data,columns=['_'.join(i.split()) for i in cancer.feature_names])
  df['y'] = cancer.target
  
  """woe计算"""
  #所有的column是
  need_columns = df.iloc[:,:-1].columns.tolist()
  data_list = []
  for i in notebook.tqdm(need_columns):
      bins = sc.woebin(df[[i,'y']],y='y')
      data_list.append(bins[i])
  
  result_woe = pd.concat(data_list)
  
  result_woe.columns = ['特征','分箱','分箱样本数','count_distr','负样本数'
                        ,'正样本数','正样本占比_当前分箱','woe_当前分箱'
                        ,'iv_当前分箱','iv值','分割点','是否特殊值']
  
  result_woe['负样本占比_当前分箱'] = result_woe['负样本数']/result_woe['分箱样本数']
  result_woe['正样本%'] = result_woe['正样本数']/df['y'].sum()
  result_woe['负样本%'] = result_woe['负样本数']/(df.shape[0] - df['y'].sum())
  
  result_woe = result_woe[['特征', '分箱', '分割点', '是否特殊值', '分箱样本数'
                        , 'count_distr', '负样本数', '正样本数', '正样本占比_当前分箱'
                        , '负样本占比_当前分箱', '正样本%','负样本%','woe_当前分箱', 'iv_当前分箱', 'iv值']]
  

  ​

• woe计算公式
  w o e = log ⁡ ( g o o d % b a d % ) woe = \log(\frac{good\%}{bad\%} ) woe=log(bad%good%​)

• iv计算公式
  i v = ∑ ( g o o d % − b a d % ) ∗ w o e iv = \sum(good\%-bad\%)*woe iv=∑(good%−bad%)∗woe