Pandas多次分组统计

Question

各位大佬，请教一个Pandas分组问题，个人感觉比较复杂。

df = pd.DataFrame({"Date":pd.date_range(start='2018-08-17 08:10:30',periods=15,freq='s',normalize=True),"Category":list('AAAAAAAAAAAABBB'),"ActionType":[1,3,2,1,4,5,3,1,3,4,5,2,1,3,2]})

首先按Category分类，类别里按Date时间排序，排序后按ActionType继续分组统计，ActionType的统计规则是，1为每个分组的开始，2为结束，但2不一定出现，示例数据的结果应该是Category分为两组A,B，其中A里又有3组，B里有一组

请教应该怎么实现或者有什么思路？

Answer 1

df = pd.DataFrame({"Date":pd.date_range(start='2018-08-17 08:10:30',periods=15,freq='s',normalize=True),"Category":list('AAAAAAAAAAAABBB'),"ActionType":[1,3,2,1,4,5,3,1,3,4,5,2,1,3,2]})
split = []
for idx,(ctg,act) in enumerate(zip(df.Category,df.ActionType)):
    if act == 1:
        split.append(idx)
        print(idx,ctg,act)
split.append(len(df.Category))
for i in range(len(split)-1):
    print(df.Date[split[i]:split[i+1]])

结果：
0 A 1
3 A 1
7 A 1
12 B 1
0 2018-08-17 00:00:00
1 2018-08-17 00:00:01
2 2018-08-17 00:00:02
Name: Date, dtype: datetime64[ns]
3 2018-08-17 00:00:03
4 2018-08-17 00:00:04
5 2018-08-17 00:00:05
6 2018-08-17 00:00:06
Name: Date, dtype: datetime64[ns]
7 2018-08-17 00:00:07
8 2018-08-17 00:00:08
9 2018-08-17 00:00:09
10 2018-08-17 00:00:10
11 2018-08-17 00:00:11
Name: Date, dtype: datetime64[ns]
12 2018-08-17 00:00:12
13 2018-08-17 00:00:13
14 2018-08-17 00:00:14
Name: Date, dtype: datetime64[ns]

Answer 2

python3

np.vsplit( df,np.where(df.ActionType==1)[0][1:])