python - Pandas DataFrame - 添加缺失月份的行

Question

嗨，我有一个如下所示的数据框

  Cust_ID   created_date   tran_date        Sales_Value  Quantity_Sold
0        1   2021-01-31    2021-01-31         126             12
1        1   2021-01-31    2021-03-31           5              3
2        1   2021-01-31    2021-06-30          20              5
3        2   2020-12-31    2021-02-28         200              5
4        2   2020-12-31    2021-03-31         250             25

我需要为每个 customer_ID 填充从创建日期到当前月份的缺失月份。基本上对于每个客户，自客户创建日期到当前日期，我应该每个月/每年有一行 因此，如果当前月份是 07/2021，则数据框应如下所示

    Cust_ID   created_date   tran_date        Sales_Value  Quantity_Sold
0        1   2021-01-31    2021-01-31         126             12
1        1   2021-01-31    2021-02-28           0              0
2        1   2021-01-31    2021-03-31           5              3
3        1   2021-01-31    2021-04-30           0              0
4        1   2021-01-31    2021-05-31           0              0
5        1   2021-01-31    2021-06-30          20              5
6        1   2021-01-31    2021-07-31           0              0
7        2   2020-12-31    2020-12-31           0              0
8        2   2020-12-31    2021-01-31           0              0
9        2   2020-12-31    2021-02-28         200              5
10       2   2020-12-31    2021-03-31         250             25
11       2   2020-12-31    2021-04-30           0              0
12       2   2020-12-31    2021-05-31           0              0
13       2   2020-12-31    2021-06-30           0              0
14       2   2020-12-31    2021-07-31           0              0

实现这一目标的最佳方法是什么？

Answer 1

使用GroupBy.apply具有自定义功能 DataFrame.reindex和 period_range将 tran_date 转换为月份:

df['created_date'] = pd.to_datetime(df['created_date'])
df['tran_date'] = pd.to_datetime(df['tran_date']).dt.to_period('m')
now = pd.to_datetime('now')
print (now)

f = lambda x: x.reindex(pd.period_range(x.name[1], now, freq='M', name='tran_date'), 
                        fill_value=0)
df = (df.set_index('tran_date')
        .groupby(['Cust_ID', 'created_date'])[['Sales_Value', 'Quantity_Sold']]
        .apply(f)
        .reset_index()
        )

df['tran_date'] = df['tran_date'].dt.to_timestamp(how='end').dt.normalize()
print (df)
    Cust_ID created_date  tran_date  Sales_Value  Quantity_Sold
0         1   2021-01-31 2021-01-31          126             12
1         1   2021-01-31 2021-02-28            0              0
2         1   2021-01-31 2021-03-31            5              3
3         1   2021-01-31 2021-04-30            0              0
4         1   2021-01-31 2021-05-31            0              0
5         1   2021-01-31 2021-06-30           20              5
6         1   2021-01-31 2021-07-31            0              0
7         2   2020-12-31 2020-12-31            0              0
8         2   2020-12-31 2021-01-31            0              0
9         2   2020-12-31 2021-02-28          200              5
10        2   2020-12-31 2021-03-31          250             25
11        2   2020-12-31 2021-04-30            0              0
12        2   2020-12-31 2021-05-31            0              0
13        2   2020-12-31 2021-06-30            0              0
14        2   2020-12-31 2021-07-31            0              0