Pandas 按 min() 条件分组

Question

我试图在本金余额低于支付金额的 5% 时获取日期变量的最小值。我希望这个按帐号提取，但我不想要一个按帐号分组的新df。

我的 df 是这样的：

| account_number | period_date | principal_balance_amt | disbursement_amt |
| -------------: | ----------- | --------------------- | ---------------- |
| 1              | 2021-01-01  | 10                    | 100              |
| 1              | 2021-02-01  | 6                     | 100              |
| 1              | 2021-03-01  | 3                     | 100              |
| 1              | 2021-04-01  | 0                     | 100              |
| 2              | 2021-01-01  | 20                    | 100              |
| 2              | 2021-02-01  | 15                    | 100              |
| 2              | 2021-03-01  | 11                    | 100              |
| 2              | 2021-04-01  | 8                     | 100              |

我已经尝试过类似的代码来让它工作，但它只是 return 无效的语法。

df['churn_date'] = df.loc[groupby('account_number').(df['principal_balance_amt'] <= 0.05 * df['disbursement_amt']), 'period_date'].min()

我希望代码创建一个如下所示的 df:

account_number	period_date	principal_balance_amt	disbursement_amt	churn_date
1	2021-01-01	10	100	2021-03-01
1	2021-02-01	6	100	2021-03-01
1	2021-03-01	3	100	2021-03-01
1	2021-04-01	0	100	2021-03-01
2	2021-01-01	20	100	nan
2	2021-02-01	15	100	nan
2	2021-03-01	11	100	nan
2	2021-04-01	8	100	nan

Answer 1

对新列使用 Series.where for replace period_date to NaN if no match and then use GroupBy.transform 和 min：

mask = (df['principal_balance_amt'] <= 0.05 * df['disbursement_amt'])
df['churn_date'] = (df.assign(new = df['period_date'].where(mask))
                      .groupby('account_number')['new']
                      .transform('min'))

print (df)
   account_number period_date  principal_balance_amt  disbursement_amt  \
0               1  2021-01-01                     10               100   
1               1  2021-02-01                      6               100   
2               1  2021-03-01                      3               100   
3               1  2021-04-01                      0               100   
4               2  2021-01-01                     20               100   
5               2  2021-02-01                     15               100   
6               2  2021-03-01                     11               100   
7               2  2021-04-01                      8               100   

  churn_date  
0 2021-03-01  
1 2021-03-01  
2 2021-03-01  
3 2021-03-01  
4        NaT  
5        NaT  
6        NaT  
7        NaT

通过 Series.map only filtered rows by boolean indexing 与聚合 min 进行映射的替代解决方案：

mask = (df['principal_balance_amt'] <= 0.05 * df['disbursement_amt'])
s = df[mask].groupby('account_number')['period_date'].min()

df['churn_date'] = df['account_number'].map(s)

Pandas 按 min() 条件分组

Pandas conditional group by min()

python

min

dataframe

pandas

pandas-groupby