Pandas：添加百分比列

Question

有 pandas DataFrame 為：

print(df)

call_id   calling_number   call_status
1          123             BUSY
2          456             BUSY
3          789             BUSY
4          123             NO_ANSWERED
5          456             NO_ANSWERED
6          789             NO_ANSWERED

在這種情況下，具有不同 call_status 的記錄（比如“錯誤”或其他，我無法預測），值可能會出現在 dataframe 中。 我需要為這樣的值動態添加一個新列。 我已經應用了 pivot_table() function 並且得到了我想要的結果：

df1 = df.pivot_table(df,index='calling_number',columns='status_code', aggfunc = 'count').fillna(0).astype('int64')

calling_number    ANSWERED  BUSY   NO_ANSWER  
123               0          1      1
456               0          1      1
789               0          1      1

現在我需要再添加一列，該列將包含具有給定 call_number 的已應答呼叫的百分比，計算為 ANSWERED 與總數的比率。 源 dataframe 'df' 可能不包含 call_status = 'ANSWERED' 的條目，因此在這種情況下，百分比列自然應該為零值。

預期結果是：

calling_number    ANSWERED  BUSY   NO_ANSWER  ANS_PERC(%)
    123               0          1      1      0
    456               0          1      1      0
    789               0          1      1      0 

Answer 1

使用crosstab ：

df1 = pd.crosstab(df['calling_number'], df['status_code'])

或者，如果需要通過count function 排除NaN ，請使用帶有添加參數pivot_table fill_value=0的 pivot_table ：

df1 = df.pivot_table(df,
               index='calling_number',
               columns='status_code', 
               aggfunc = 'count', 
               fill_value=0)

然后對於比率除以每行的總和值：

df1 = df1.div(df1.sum(axis=1), axis=0)
print (df1)
                ANSWERED      BUSY  NO_ANSWER
calling_number                               
123             0.333333  0.333333   0.333333
456             0.333333  0.333333   0.333333
789             0.333333  0.333333   0.333333

編輯：為了添加可能不存在的某些類別，請使用DataFrame.reindex ：

df1 = (pd.crosstab(df['calling_number'], df['call_status'])
         .reindex(columns=['ANSWERED','BUSY','NO_ANSWERED'], fill_value=0))

df1['ANS_PERC(%)'] = df1['ANSWERED'].div(df1['ANSWERED'].sum()).fillna(0)
print (df1)
call_status     ANSWERED  BUSY  NO_ANSWERED  ANS_PERC(%)
calling_number                                          
123                    0     1            1          0.0
456                    0     1            1          0.0
789                    0     1            1          0.0

如果需要每行總數：

df1['ANS_PERC(%)'] = df1['ANSWERED'].div(df1.sum(axis=1))
print (df1)
call_status     ANSWERED  BUSY  NO_ANSWERED  ANS_PERC(%)
calling_number                                          
123                    0     1            1          0.0
456                    0     1            1          0.0
789                    0     1            1          0.0

編輯1：

將一些錯誤值替換為ERROR的解決方案：

print (df)
   call_id  calling_number  call_status
0        1             123          ttt
1        2             456         BUSY
2        3             789         BUSY
3        4             123  NO_ANSWERED
4        5             456  NO_ANSWERED
5        6             789  NO_ANSWERED

L = ['ANSWERED', 'BUSY', 'NO_ANSWERED']
df['call_status'] = df['call_status'].where(df['call_status'].isin(L), 'ERROR')
print (df)
0        1             123        ERROR
1        2             456         BUSY
2        3             789         BUSY
3        4             123  NO_ANSWERED
4        5             456  NO_ANSWERED
5        6             789  NO_ANSWERED
df1 = (pd.crosstab(df['calling_number'], df['call_status'])
         .reindex(columns=L + ['ERROR'], fill_value=0))

df1['ANS_PERC(%)'] = df1['ANSWERED'].div(df1.sum(axis=1))
print (df1)
call_status     ANSWERED  BUSY  NO_ANSWERED  ERROR  ANS_PERC(%)
calling_number                                                 
123                    0     0            1      1          0.0
456                    0     1            1      0          0.0
789                    0     1            1      0          0.0

Answer 2

我喜歡 cross_tab 的想法，但我是列操作的粉絲，因此很容易參考：

    # define a function to capture all the other call_statuses into one bucket 
def tester(x):
    if x not in ['ANSWERED', 'BUSY', 'NO_ANSWERED']:
        return 'OTHER' 
    else:
        return x
    
#capture the simplified status in a new column
df['refined_status'] = df['call_status'].apply(tester)


#Do the pivot (or cross tab) to capture the sums:
df1= df.pivot_table(values="call_id", index = 'calling_number', columns='refined_status', aggfunc='count')

#Apply a division to get the percentages:
df1["TOTAL"] = df1[['ANSWERED', 'BUSY', 'NO_ANSWERED', 'OTHER']].sum(axis=1)
df1["ANS_PERC"] = df1["ANSWERED"]/df1.TOTAL * 100

print(df1)