""" 分箱邏輯：1.類別型特征： 1）類別數在5個以下，可以直接根據類別來分箱 (binning_cate) 2）類別數在5個以上，建議做降基處理，再根據降基后的類別做分箱2.數值型特征： 1）離散型數值特征（特征value的變動幅度較小）：若特征value的非重復計數在5個以下，可以直接根據非重復計數值來分箱(binning_cate)若特征value的非重復計數在5個以上，建議根據業務解釋或者數據分布做自定義分箱(binning_self) 2）連續型數值特征（特征value的變動幅度較大）：可以用卡方分箱或自定義分箱。(binning_num,binning_self)PS:一些特征用卡方分可能會報錯，建議這些特征改為手動自定義分箱3.特征有缺失： 1）缺失率在5%以下，可以先對缺失做填充處理再分箱(binning_num) 2）缺失率在5%以上，建議將缺失當作一個類別來分箱(binning_sparse_col)4.稀疏特征分箱 建議將稀疏值（一般為0）單獨分為一箱，剩下的值做卡方或者自定義分箱(binning_sparse_col) """ def binning_cate(df, col, target):"""df:數據集col:輸入的特征target:好壞標記的字段名return:bin_df :特征的評估結果"""total = df[target].count()bad = df[target].sum()good = total - badd1 = df.groupby([col], as_index=True)d2 = pd.DataFrame()d2['樣本數'] = d1[target].count()d2['黑樣本數'] = d1[target].sum()d2['白樣本數'] = d2['樣本數'] - d2['黑樣本數']d2['逾期用戶占比'] = d2['黑樣本數'] / d2['樣本數']d2['badattr'] = d2['黑樣本數'] / badd2['goodattr'] = d2['白樣本數'] / goodd2['WOE'] = np.log(d2['badattr'] / d2['goodattr'])d2['bin_iv'] = (d2['badattr'] - d2['goodattr']) * d2['WOE']d2['IV'] = d2['bin_iv'].sum()bin_df = d2.reset_index()bin_df.drop(['badattr', 'goodattr', 'bin_iv'], axis=1, inplace=True)bin_df.rename(columns={col: '分箱結果'}, inplace=True)bin_df['特征名'] = colbin_df = pd.concat([bin_df['特征名'], bin_df.iloc[:, :-1]], axis=1)return bin_dfdef binning_self(df, col, target, cut=None, right_border=True):"""df:數據集col:輸入的特征target:好壞標記的字段名cut:總定義劃分區間的listright_border：設定左開右閉、左閉右開return:bin_df :特征的評估結果"""total = df[target].count()bad = df[target].sum()good = total - badbucket = pd.cut(df[col], cut, right=right_border)d1 = df.groupby(bucket)d2 = pd.DataFrame()d2['樣本數'] = d1[target].count()d2['黑樣本數'] = d1[target].sum()d2['白樣本數'] = d2['樣本數'] - d2['黑樣本數']d2['逾期用戶占比'] = d2['黑樣本數'] / d2['樣本數']d2['badattr'] = d2['黑樣本數'] / badd2['goodattr'] = d2['白樣本數'] / goodd2['WOE'] = np.log(d2['badattr'] / d2['goodattr'])d2['bin_iv'] = (d2['badattr'] - d2['goodattr']) * d2['WOE']d2['IV'] = d2['bin_iv'].sum()bin_df = d2.reset_index()bin_df.drop(['badattr', 'goodattr', 'bin_iv'], axis=1, inplace=True)bin_df.rename(columns={col: '分箱結果'}, inplace=True)bin_df['特征名'] = colbin_df = pd.concat([bin_df['特征名'], bin_df.iloc[:, :-1]], axis=1)ks, precision, tpr, fpr = cal_ks(df, col, target)bin_df['準確率'] = precisionbin_df['召回率'] = tprbin_df['打擾率'] = fprbin_df['KS'] = ksreturn bin_dfdef binning_num(df, target, col, max_bin=None, min_binpct=None):"""df:數據集col:輸入的特征target:好壞標記的字段名max_bin:最大的分箱個數min_binpct:區間內樣本所占總體的最小比return:bin_df :特征的評估結果"""total = df[target].count()bad = df[target].sum()good = total - badinf = float('inf')ninf = float('-inf')cut = ChiMerge(df, col, target, max_bin=max_bin, min_binpct=min_binpct)cut.insert(0, ninf)cut.append(inf)bucket = pd.cut(df[col], cut)d1 = df.groupby(bucket)d2 = pd.DataFrame()d2['樣本數'] = d1[target].count()d2['黑樣本數'] = d1[target].sum()d2['白樣本數'] = d2['樣本數'] - d2['黑樣本數']d2['逾期用戶占比'] = d2['黑樣本數'] / d2['樣本數']d2['badattr'] = d2['黑樣本數'] / badd2['goodattr'] = d2['白樣本數'] / goodd2['WOE'] = np.log(d2['badattr'] / d2['goodattr'])d2['bin_iv'] = (d2['badattr'] - d2['goodattr']) * d2['WOE']d2['IV'] = d2['bin_iv'].sum()bin_df = d2.reset_index()bin_df.drop(['badattr', 'goodattr', 'bin_iv'], axis=1, inplace=True)bin_df.rename(columns={col: '分箱結果'}, inplace=True)bin_df['特征名'] = colbin_df = pd.concat([bin_df['特征名'], bin_df.iloc[:, :-1]], axis=1)ks, precision, tpr, fpr = cal_ks(df, col, target)bin_df['準確率'] = precisionbin_df['召回率'] = tprbin_df['打擾率'] = fprbin_df['KS'] = ksreturn bin_dfdef binning_sparse_col(df, target, col, max_bin=None, min_binpct=None, sparse_value=None):"""df:數據集col:輸入的特征target:好壞標記的字段名max_bin:最大的分箱個數min_binpct:區間內樣本所占總體的最小比sparse_value:單獨分為一箱的value值return:bin_df :特征的評估結果"""total = df[target].count()bad = df[target].sum()good = total - bad# 對稀疏值0值或者缺失值單獨分箱temp1 = df[df[col] == sparse_value]temp2 = df[~(df[col] == sparse_value)]bucket_sparse = pd.cut(temp1[col], [float('-inf'), sparse_value])group1 = temp1.groupby(bucket_sparse)bin_df1 = pd.DataFrame()bin_df1['樣本數'] = group1[target].count()bin_df1['黑樣本數'] = group1[target].sum()bin_df1['白樣本數'] = bin_df1['樣本數'] - bin_df1['黑樣本數']bin_df1['逾期用戶占比'] = bin_df1['黑樣本數'] / bin_df1['樣本數']bin_df1['badattr'] = bin_df1['黑樣本數'] / badbin_df1['goodattr'] = bin_df1['白樣本數'] / goodbin_df1['WOE'] = np.log(bin_df1['badattr'] / bin_df1['goodattr'])bin_df1['bin_iv'] = (bin_df1['badattr'] - bin_df1['goodattr']) * bin_df1['WOE']bin_df1 = bin_df1.reset_index()# 對剩余部分做卡方分箱cut = ChiMerge(temp2, col, target, max_bin=max_bin, min_binpct=min_binpct)cut.insert(0, sparse_value)cut.append(float('inf'))bucket = pd.cut(temp2[col], cut)group2 = temp2.groupby(bucket)bin_df2 = pd.DataFrame()bin_df2['樣本數'] = group2[target].count()bin_df2['黑樣本數'] = group2[target].sum()bin_df2['白樣本數'] = bin_df2['樣本數'] - bin_df2['黑樣本數']bin_df2['逾期用戶占比'] = bin_df2['黑樣本數'] / bin_df2['樣本數']bin_df2['badattr'] = bin_df2['黑樣本數'] / badbin_df2['goodattr'] = bin_df2['白樣本數'] / goodbin_df2['WOE'] = np.log(bin_df2['badattr'] / bin_df2['goodattr'])bin_df2['bin_iv'] = (bin_df2['badattr'] - bin_df2['goodattr']) * bin_df2['WOE']bin_df2 = bin_df2.reset_index()# 合并分箱結果bin_df = pd.concat([bin_df1, bin_df2], axis=0)bin_df['IV'] = bin_df['bin_iv'].sum().round(3)bin_df.drop(['badattr', 'goodattr', 'bin_iv'], axis=1, inplace=True)bin_df.rename(columns={col: '分箱結果'}, inplace=True)bin_df['特征名'] = colbin_df = pd.concat([bin_df['特征名'], bin_df.iloc[:, :-1]], axis=1)ks, precision, tpr, fpr = cal_ks(df, col, target)bin_df['準確率'] = precisionbin_df['召回率'] = tprbin_df['打擾率'] = fprbin_df['KS'] = ksreturn bin_df

---

??對數據分析、機器學習、數據科學、金融風控等感興趣的小伙伴，需要數據集、代碼、行業報告等各類學習資料，可添加微信：wu805686220（記得要備注喔！），也可關注微信公眾號：風控圏子（別打錯字，是圏子，不是圈子，算了直接復制吧！）

關注公眾號后，可聯系圈子助手加入如下社群：

• 機器學習風控討論群（微信群）
• 反欺詐討論群（微信群）
• python學習交流群（微信群）
• 研習社資料（qq群：102755159）（干貨、資料、項目、代碼、報告、課件）

相互學習，共同成長。

總結

以上是生活随笔為你收集整理的【数据处理】python变量分箱常见手法：分类型、数值型、卡方、自定义的全部內容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網站內容還不錯，歡迎將生活随笔推薦給好友。