我有一个Dataframe将其更改为时间序列。日期范围从2013年到2017年。我想按小时分组所有数据。 例如，所有星期一一起表示每小时，然后是所有星期二。最后我会有168（24 * 7）行。 这样做的最佳方式是什么？
重新取样后我有这个样本：
2017-01-17 00:00:00 NaN
2017-01-17 01:00:00 NaN
2017-01-17 02:00:00 NaN
2017-01-17 03:00:00 NaN
2017-01-17 04:00:00 1.0
2017-01-17 05:00:00 NaN
2017-01-17 06:00:00 NaN
2017-01-17 07:00:00 NaN
2017-01-17 08:00:00 NaN
2017-01-17 09:00:00 1.0
2017-01-17 10:00:00 3.0
2017-01-17 11:00:00 3.0
2017-01-17 12:00:00 3.0
2017-01-17 13:00:00 5.0
2017-01-17 14:00:00 2.0
2017-01-17 15:00:00 1.0
2017-01-17 16:00:00 2.0
2017-01-17 17:00:00 1.0
2017-01-17 18:00:00 1.0
2017-01-17 19:00:00 1.0
2017-01-17 20:00:00 NaN
2017-01-17 21:00:00 NaN
2017-01-17 22:00:00 NaN
2017-01-17 23:00:00 NaN
2017-01-24 10:00:00 14.0
2017-01-24 11:00:00 14.0
2017-01-24 12:00:00 5.0
2017-01-24 13:00:00 21.0
2017-01-24 14:00:00 14.0
2017-01-24 15:00:00 7.0
2017-01-24 16:00:00 9.0
2017-01-24 17:00:00 2.0
2017-01-24 18:00:00 1.0
2017-01-24 19:00:00 NaN
2017-01-24 20:00:00 NaN
2017-01-24 21:00:00 2.0
我希望有类似的东西：
(count sum)
Monday: 00:00 xx
01:00 xx
...
23:00 xx
Tuesday: 00:00 xx
01:00 xx
...
23:00 xx
答案 0 :(得分：5)
我认为你可以groupby和dayofweek hour聚合某些功能，例如sum：
np.random.seed(100)
start = pd.to_datetime('2013-02-24 04:00:00')
rng = pd.date_range(start, periods=100, freq='3H')
#DataFrame has DatetimeIndex
df = pd.DataFrame({'a': np.random.randint(10, size=100)}, index=rng)
print (df)
a
2013-02-24 04:00:00 8
2013-02-24 07:00:00 8
2013-02-24 10:00:00 3
2013-02-24 13:00:00 7
2013-02-24 16:00:00 7
2013-02-24 19:00:00 0
2013-02-24 22:00:00 4
2013-02-25 01:00:00 2
2013-02-25 04:00:00 5
...
...
print (df.index.weekday_name)
['Sunday' 'Sunday' 'Sunday' 'Sunday' 'Sunday' 'Sunday' 'Sunday' 'Monday'
'Monday' 'Monday' 'Monday' 'Monday' 'Monday' 'Monday' 'Monday' 'Tuesday'
'Tuesday' 'Tuesday' 'Tuesday' 'Tuesday' 'Tuesday' 'Tuesday' 'Tuesday'
'Wednesday' 'Wednesday' 'Wednesday' 'Wednesday' 'Wednesday' 'Wednesday'
'Wednesday' 'Wednesday' 'Thursday' 'Thursday' 'Thursday' 'Thursday'
'Thursday' 'Thursday' 'Thursday' 'Thursday' 'Friday' 'Friday' 'Friday'
'Friday' 'Friday' 'Friday' 'Friday' 'Friday' 'Saturday' 'Saturday'
'Saturday' 'Saturday' 'Saturday' 'Saturday' 'Saturday' 'Saturday' 'Sunday'
'Sunday' 'Sunday' 'Sunday' 'Sunday' 'Sunday' 'Sunday' 'Sunday' 'Monday'
'Monday' 'Monday' 'Monday' 'Monday' 'Monday' 'Monday' 'Monday' 'Tuesday'
'Tuesday' 'Tuesday' 'Tuesday' 'Tuesday' 'Tuesday' 'Tuesday' 'Tuesday'
'Wednesday' 'Wednesday' 'Wednesday' 'Wednesday' 'Wednesday' 'Wednesday'
'Wednesday' 'Wednesday' 'Thursday' 'Thursday' 'Thursday' 'Thursday'
'Thursday' 'Thursday' 'Thursday' 'Thursday' 'Friday' 'Friday' 'Friday'
'Friday' 'Friday']
print (df.index.hour)
[ 4 7 10 13 16 19 22 1 4 7 10 13 16 19 22 1 4 7 10 13 16 19 22 1 4
7 10 13 16 19 22 1 4 7 10 13 16 19 22 1 4 7 10 13 16 19 22 1 4 7
10 13 16 19 22 1 4 7 10 13 16 19 22 1 4 7 10 13 16 19 22 1 4 7 10
13 16 19 22 1 4 7 10 13 16 19 22 1 4 7 10 13 16 19 22 1 4 7 10 13]
print (df.groupby([df.index.weekday_name, df.index.hour])['a'].sum())
Friday 1 13
4 10
7 6
10 13
13 11
16 2
19 0
22 8
Monday 1 6
4 12
7 8
10 5
13 11
...
...
如果DataFrame有date列：
np.random.seed(100)
start = pd.to_datetime('2013-02-24 04:00:00')
rng = pd.date_range(start, periods=100, freq='3H')
df = pd.DataFrame({'date': rng, 'a': np.random.randint(10, size=100)})
print (df)
a date
0 8 2013-02-24 04:00:00
1 8 2013-02-24 07:00:00
2 3 2013-02-24 10:00:00
3 7 2013-02-24 13:00:00
4 7 2013-02-24 16:00:00
5 0 2013-02-24 19:00:00
6 4 2013-02-24 22:00:00
7 2 2013-02-25 01:00:00
8 5 2013-02-25 04:00:00
print (df.groupby([df.date.dt.weekday_name, df.date.dt.hour])['a'].sum())
date date
Friday 1 13
4 10
7 6
10 13
13 11
16 2
19 0
22 8
Monday 1 6
4 12
7 8
10 5
13 11
如果Series与DatetimeIndex：
s = pd.Series(np.random.randint(10, size=100), index=rng)
print (s)
2013-02-24 04:00:00 8
2013-02-24 07:00:00 8
2013-02-24 10:00:00 3
2013-02-24 13:00:00 7
2013-02-24 16:00:00 7
2013-02-24 19:00:00 0
2013-02-24 22:00:00 4
2013-02-25 01:00:00 2
2013-02-25 04:00:00 5
2013-02-25 07:00:00 2
2013-02-25 10:00:00 2
2013-02-25 13:00:00 2
print (s.groupby([s.index.weekday_name, s.index.hour]).sum())
Friday 1 13
4 10
7 6
10 13
13 11
16 2
19 0
22 8
Monday 1 6
4 12
7 8
10 5
13 11
最后为DataFrame添加reset_index()：
df = s.groupby([s.index.weekday_name, s.index.hour]).sum().reset_index()
df.columns = ['days','hours','val']
print (df)
days hours val
0 Friday 1 13
1 Friday 4 10
2 Friday 7 6
3 Friday 10 13
4 Friday 13 11
5 Friday 16 2
6 Friday 19 0
7 Friday 22 8
8 Monday 1 6
9 Monday 4 12
10 Monday 7 8
11 Monday 10 5
12 Monday 13 11
通过评论编辑：
print (s)
2017-01-24 10:00:00 14.0
2017-01-24 11:00:00 14.0
2017-01-24 12:00:00 5.0
2017-01-24 13:00:00 21.0
2017-01-24 14:00:00 14.0
2017-01-24 15:00:00 7.0
2017-01-24 16:00:00 9.0
2017-01-24 17:00:00 2.0
2017-01-24 18:00:00 1.0
2017-01-24 19:00:00 NaN
2017-01-24 20:00:00 NaN
2017-01-24 21:00:00 2.0
Name: a, dtype: float64
df = s.groupby([s.index.weekday_name, s.index.hour]).sum().reset_index()
df.columns = ['days','hours','val']
print (df)
days hours val
0 Tuesday 10 14.0
1 Tuesday 11 14.0
2 Tuesday 12 5.0
3 Tuesday 13 21.0
4 Tuesday 14 14.0
5 Tuesday 15 7.0
6 Tuesday 16 9.0
7 Tuesday 17 2.0
8 Tuesday 18 1.0
9 Tuesday 19 NaN
10 Tuesday 20 NaN
11 Tuesday 21 2.0