在python中只对那些彼此相邻的相同组进行分组的方法是什么？

Question

输入：

   Column1    Event   Time
0   type1     A       2020-11-01 05:26:12
1   type1     A       2020-11-01 05:26:17
2   type1     A       2020-11-01 05:26:22
3   type1     B       2020-11-01 05:26:32
4   type1     B       2020-11-01 05:26:37
5   type1     A       2020-11-01 05:26:42
6   type1     A       2020-11-01 05:26:47
7   type1     C       2020-11-01 05:26:52
8   type1     C       2020-11-01 05:27:02
9   type1     B       2020-11-01 05:27:07
10  type1     B       2020-11-01 05:27:15

期望的输出：

    Column1   Event   Min(Time)             Max(Time)            EventEndTime

0   type1     A       2020-11-01 05:26:12   2020-11-01 05:26:22  2020-11-01 05:26:32
1   type1     B       2020-11-01 05:26:32   2020-11-01 05:26:37  2020-11-01 05:26:42
2   type1     A       2020-11-01 05:26:42   2020-11-01 05:26:47  2020-11-01 05:26:52
3   type1     C       2020-11-01 05:26:52   2020-11-01 05:27:02  2020-11-01 05:27:07
4   type1     B       2020-11-01 05:27:07   2020-11-01 05:27:15

我已尝试但无法仅对相邻的相同事件进行分组，我最终通过以下代码行获得了总体最小值和最大值：

 data.groupby(['Column1','Event']).agg({'Time': ['min','max']}) 

Answer 1

您可以使用辅助系列将移位值与累积总和与g 进行比较，并且聚合用于字典中的命名聚合，因此用于解包**，最后由DataFrame.reset_index 删除第三级DataFrame.reset_index drop=True:

g = data['Event'].ne(data['Event'].shift()).cumsum()

d = {'Min(Time)': ('Time', 'min'), 'Max(Time)': ('Time', 'max')}
df = (data.groupby(['Column1','Event', g], sort=False)
          .agg(**d)
          .reset_index(level=2, drop=True)
          .reset_index())
print (df)
  Column1 Event           Min(Time)           Max(Time)
0   type1     A 2020-11-01 05:26:12 2020-11-01 05:26:22
1   type1     B 2020-11-01 05:26:32 2020-11-01 05:26:37
2   type1     A 2020-11-01 05:26:42 2020-11-01 05:26:47
3   type1     C 2020-11-01 05:26:52 2020-11-01 05:27:02
4   type1     B 2020-11-01 05:27:07 2020-11-01 05:27:15

如果要创建辅助列：

data['groups'] = data['Event'].ne(data['Event'].shift()).cumsum()
print (data)
   Column1 Event                Time  groups
0    type1     A 2020-11-01 05:26:12       1
1    type1     A 2020-11-01 05:26:17       1
2    type1     A 2020-11-01 05:26:22       1
3    type1     B 2020-11-01 05:26:32       2
4    type1     B 2020-11-01 05:26:37       2
5    type1     A 2020-11-01 05:26:42       3
6    type1     A 2020-11-01 05:26:47       3
7    type1     C 2020-11-01 05:26:52       4
8    type1     C 2020-11-01 05:27:02       4
9    type1     B 2020-11-01 05:27:07       5
10   type1     B 2020-11-01 05:27:15       5

d = {'Min(Time)': ('Time', 'min'), 'Max(Time)': ('Time', 'max')}
df = (data.groupby(['Column1','Event', 'groups'], sort=False)
          .agg(**d)
          .reset_index(level=2, drop=True)
          .reset_index())
print (df)
  Column1 Event           Min(Time)           Max(Time)
0   type1     A 2020-11-01 05:26:12 2020-11-01 05:26:22
1   type1     B 2020-11-01 05:26:32 2020-11-01 05:26:37
2   type1     A 2020-11-01 05:26:42 2020-11-01 05:26:47
3   type1     C 2020-11-01 05:26:52 2020-11-01 05:27:02
4   type1     B 2020-11-01 05:27:07 2020-11-01 05:27:15

编辑：如果需要每组的下一个事件：

df['EventEndTime'] = df.groupby('Column1')['Min(Time)'].shift(-1)

#next Event not per groups
#df['EventEndTime'] = df['Min(Time)'].shift(-1)

print (df)
  Column1 Event           Min(Time)           Max(Time)        EventEndTime
0   type1     A 2020-11-01 05:26:12 2020-11-01 05:26:22 2020-11-01 05:26:32
1   type1     B 2020-11-01 05:26:32 2020-11-01 05:26:37 2020-11-01 05:26:42
2   type1     A 2020-11-01 05:26:42 2020-11-01 05:26:47 2020-11-01 05:26:52
3   type1     C 2020-11-01 05:26:52 2020-11-01 05:27:02 2020-11-01 05:27:07
4   type1     B 2020-11-01 05:27:07 2020-11-01 05:27:15                 NaT