vif-方差膨胀因子计算

vif-方差膨胀因子

• 使用statsmodels中的variance_inflation_factor，数据集使用乳腺癌数据集

  import pandas as pd
  import numpy as np
  from sklearn.datasets import load_breast_cancer
  from tqdm import notebook
  from statsmodels.stats.outliers_influence import variance_inflation_factor
  from statsmodels.tools import add_constant
  
  cancer = load_breast_cancer()
  df = pd.DataFrame(cancer.data,columns=['_'.join(i.split()) for i in cancer.feature_names])
  df['y'] = cancer.target
  
  def calc_vif(df,desc_dict=None):
      """
      df: DataFrame
      desc_dict: 字段中文释义字典
      
      return: vif DataFrame
      """
      X=add_constant(df)
      vif_list = []
      for i in notebook.tqdm(range(X.shape[1])):
          vif = variance_inflation_factor(X.values,i)
          vif_list.append(vif)
      s = pd.DataFrame(vif_list,index=X.columns)
      s = s.reset_index()
      s.columns = ['feature','vif']
      if desc_dict:
          s['中文释义'] = s.feature.map(desc_dict)
          s = s[['feature','中文释义','vif']]
      return s
    
  str_list = df.select_dtypes(include='object').columns.tolist() # string类型
  flo_list = df.select_dtypes(exclude='object').columns.tolist() # 连续型
  
  # vif 不能计算string类型
  df_vif = df[flo_list]
  
  # vif 计算不能有空值
  for i in notebook.tqdm(flo_list):
      df_vif[i] = df_vif[i].fillna(0)
  
  vif_data = calc_vif(df_vif)
  vif_data