def get_kind(x: pd.Series, diff_limit: int = 10):
    x = x.astype('str')
    x = x.str.extract(r'(^(\-|)(?=.*\d)\d*(?:\.\d*)?$)')[0]
    x.dropna(inplace=True)
    if x.nunique() > diff_limit:
        kind = 'numeric'
    else:
        kind = 'categorical'
    return kind


class feature_select(BaseEstimator, TransformerMixin):
    def __init__(self,
              num_list: list = None,
              cate_list: list = None,
              num_method: str='sys',
              cate_method: str='sys',
              diff_num: int = 10,
              pos_label: str = 1,
              show_df: bool = False,
              ):
        self.num_list = num_list
        self.cate_list = cate_list
        self.num_method= num_method
        self.cate_method= cate_method
        self.diff_num = diff_num
        self.pos_label = pos_label
        self.show_df = show_df
        self.selete_list = []

    def fit (self, X, y=None):
        X=X.copy()
        from scipy import stats
        if self.num_list is None:
            self.num_list = []
            for col in X.columns:
                kind = get_kind(x=X[col], diff_limit=self.diff_num)
                if kind == 'numeric':
                    self.num_list.append(col)

        if self.cate_list is None:
            self.cate_list = []
            for col in X.columns:
                kind = get_kind(x=X[col], diff_limit=self.diff_num)
                if kind == 'categorical':
                    self.cate_list.append(col)

        X['y']=y
        yes=X[X['y'] == self.pos_label]
        yes.reset_index(drop=True,inplace=True)
        no=X[X['y'] != self.pos_label]
        no.reset_index(drop=True,inplace=True)
        del X['y']

        sys_cate_list, kf_list, kf_p_list =[],[],[]
        sys_num_list, t_list, p_value_list, anova_f_list, anova_p_list=[],[],[],[],[]

        if self.cate_method == 'sys' or self.show_df is True:
            for obj in self.cate_list:
                value_list = list(X[obj].unique())
                value_sum = 0
                for value in value_list:
                    support_yes = (yes[yes[obj] == value].shape[0] +1)/ ( yes.shape[0]+1)
                    support_no = (no[no[obj] == value].shape[0] +1) / (no.shape[0]+1)
                    confidence_yes = support_yes / (support_yes + support_no)
                    value_sum += abs(2 * confidence_yes - 1) * (X[X[obj] == value].shape[0] / X.shape[0])
                sys_cate_list.append(value_sum)
                if value_sum >=0.1:
                    self.select_list.append(obj)

        if self.cate_method == 'kf' or self.show_df is True:
            for obj in self.cate_list:
                df_obj=pd.get_dummies(X[obj],prefix=obj)
                df_obj['result']=y
                df_obj=df_obj,groupby('result').sum()
                obs = df_obj.values
                kf = stats.chi2_contingency(obs)
                '''
                chi2: The test statistic
                p: p-value
                dof: Degrees of freedom
                expected: The expected frequencies, based on the marginal sums of the table.
                '''
                chi2, p, dof, expect = kf
                kf_list.append(chi2)
                kf_p_list.append(p)

                if p < 0.05:
                    self.select_list.append(obj)

        if self.num_method == 'sys' or self.show_df is True:
            for num in self.num_list:
                mean_c1 = no[num].mean()
                std_c1 = no[num].std()
                mean_c2 = yes[num].mean()
                std_c2 = yes[num].std()
                value_sum=abs(mean_c1 - mean_c2) / (std_c1 + std_c2) * 2
                sys_num_list.append(value_sum)
                if value_sum >=0.1:
                    self.select_list.append(num)

        if self.num_method == 't' or self.show_df is True:
            for num in num_list:
                t_t, t_p = stats.ttest_ind(yes[num], no[num], equal_var=False, nan_policy='omit')  # 'omit'忽略nan值执行计算
                t_list.append(t_t)
                p_value_list.append(t_p)
                if t_p < 0.05:
                    self.select_list.append(num)

        if self.num_method == 'anova' or self.show_df is True:
            for num in num_list:
                anova_f, anova_p = stats.f_oneway(yes[num], no[num])
                anova_f_list.append(anova_f)
                anova_p_list.append(anova_p)
                if anova_p < 0.05:
                    self.select_list.append(num)

        if self.show_df is True:
            dic1 = {'categorical': self.cate_list, 'importance_': sys_cate_list, 'Kf-Value': kf_list, 'Kf_P-Value': kf_p_list, }
            df = pd.DataFrame(dic1, columns=['categorical', 'importance_', 'Kf-Value', 'Kf_P-Value'])
            df.sort_values(by='Kf_P-Value', inplace=True)
            print(df,'\n')

            dic2 = {'numeric':self. num_list, 'importance_': sys_num_list, 'T-Value': t_list, 'P-value': p_value_list, 'Anova-F-Value': anova_f_list, 'Anova-P-value': anova_p_list}
            df = pd.DataFrame(dic2, columns=['numeric', 'importance_', 'T-Value', 'P-value', 'Anova-F-Value', 'Anova-P-value'])
            df.sort_values(by='Anova-P-value', inplace=True)
            print(df,'\n')

        self.select_list= list(set(self.select_list))
        print('After select attr:', self.select_list)
        return self
    
    def transform(self,X):
        X=X.copy()
        print('attr select success!')
        return X[self.select_list]
        
        
fs=feature_select(pos_label=’yes’,show_df=True)
x_train=fs.fit_transform(x_train, y_train)

fit的时候报错，提示如下：

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)in1 fs=feature_select(pos_label='1',show_df=True)
----> 2 x_=fs.fit_transform(x_, y_)

C:\ProgramData\Anaconda3\lib\site-packages\sklearn\base.py in fit_transform(self, X, y, **fit_params)
    700         else:
    701             # fit method of arity 2 (supervised transformation)
--> 702             return self.fit(X, y, **fit_params).transform(X)
    703 
    704in fit(self, X, y)
     56                 sys_cate_list.append(value_sum)
     57                 if value_sum >=0.1:
---> 58                     self.select_list.append(obj)
     59 
     60         if self.cate_method == 'kf' or self.show_df is True:

AttributeError: 'feature_select' object has no attribute 'select_list'