如何使col2成為每組DateTime中第二個col1=='x'的總和？

我想要一個pandas dataframe專欄

• 計算在through'outcome2'中觀察到的次數'value''datetime'
• 從第二次觀察開始'outcome2'
• 每'ID'或df.index

import pandas as pd
from io import StringIO
import datetime 

txt= """
ID,datetime,value
A,12/10/2022 10:00:00,outcome1
A,12/10/2022 11:15:10,outcome2
A,14/10/2022 15:30:30,outcome1
B,11/10/2022 11:30:22,outcome1
B,15/10/2022 22:44:11,outcome2
B,15/10/2022 23:30:22,outcome3
B,15/10/2022 23:31:11,outcome2
"""

df = pd.read_csv(StringIO(txt),\
             parse_dates=[1],\
             dayfirst=True)\
             .assign(id_index= lambda x_df: x_df\
             .groupby('ID', sort=False).ngroup())\
             .set_index("id_index")\
             .rename_axis(index=None)

df = df.assign(value_test = lambda df: df['value']=='outcome2',\
               value_cumsum= lambda df: df.groupby('ID', sort=False)['value_test'].cumsum())

  ID            datetime     value value_test   value_cumsum
0  A 2022-10-12 10:00:00  outcome1      False              0
0  A 2022-10-12 11:15:10  outcome2       True              1
0  A 2022-10-14 15:30:30  outcome1      False              1
1  B 2022-10-11 11:30:22  outcome1      False              0
1  B 2022-10-15 22:44:11  outcome2       True              1
1  B 2022-10-15 23:30:22  outcome3      False              1
1  B 2022-10-15 23:31:11  outcome2       True              2

我嘗試將第三個變數分配給在函式中df使用 if 陳述句。lambda它以其他人經歷過的方式失敗1：

df = df.assign(value_test = lambda df: df['value']=='outcome2',\
               value_cumsum = lambda df: df.groupby('ID', sort=False)['value_test'].cumsum(),\
               outcome2 = lambda df: 0 if df[df[value_cumsum]==1] or df[df[value_cumsum]==0]\
               else df[value_cumsum]-1 if df[df[value_cumsum] > 1]

ValueError: The truth value of a DataFrame is ambiguous. Use a.empty, a.bool(), a.item(), a.any() or a.all().

我只需要從每組的第二次觀察開始的計數的累積總和（運行總數）'outcome2'。 *'value''outcome2'

請問有什么建議嗎？

是否可以使用lambdavalue_testvalue_cumsum或不使用中間步驟或？

所需的df

  ID            datetime     value outcome2
0  A 2022-10-12 10:00:00  outcome1        0
0  A 2022-10-12 11:15:10  outcome2        0
0  A 2022-10-14 15:30:30  outcome1        0
1  B 2022-10-11 11:30:22  outcome1        0
1  B 2022-10-15 22:44:11  outcome2        0
1  B 2022-10-15 23:30:22  outcome3        0
1  B 2022-10-15 23:31:11  outcome2        1

uj5u.com熱心網友回復：

您可以使用：

df['value_cumsum'] = (df.groupby('ID')['value_test']
                      .cumsum().sub(1).where(df['value_test'], 0)
                      )

或者，如果您還想標記 False：

df['value_cumsum'] = (df.groupby('ID')['value_test']
                      .cumsum().sub(1).clip(lower=0)
                      )

輸出：

  ID            datetime     value  value_test  value_cumsum
0  A 2022-10-12 10:00:00  outcome1       False             0
0  A 2022-10-12 11:15:10  outcome2        True             0
0  A 2022-10-14 15:30:30  outcome1       False             0
1  B 2022-10-11 11:30:22  outcome1       False             0
1  B 2022-10-15 22:44:11  outcome2        True             0
1  B 2022-10-15 23:30:22  outcome3       False             0
1  B 2022-10-15 23:31:11  outcome2        True             1

無中間體：

df['value_cumsum'] = (df['value'].eq('outcome2')
                      .groupby(df['ID'])
                      .cumsum().sub(1).clip(lower=0)
                     )

輸出：

  ID            datetime     value  value_cumsum
0  A 2022-10-12 10:00:00  outcome1             0
0  A 2022-10-12 11:15:10  outcome2             0
0  A 2022-10-14 15:30:30  outcome1             0
1  B 2022-10-11 11:30:22  outcome1             0
1  B 2022-10-15 22:44:11  outcome2             0
1  B 2022-10-15 23:30:22  outcome3             0
1  B 2022-10-15 23:31:11  outcome2             1

標籤：Python熊猫数据框拉姆达cumsum

上一篇：為驗證集創建滯后

下一篇：使用正則運算式提取日期和時間，然后將其轉換為日期時間