python - 如何将 pandas.DataFrame (一周)重新采样为平均日

Question

我有几天(甚至几周)每天在完全相同的时间间隔获取的数据，并且想要计算平均日时间曲线。到目前为止，我尝试了每日平均值，但每天都有一个平均值...我需要的是每个可用时间的所有可用天数的平均值。很可能真的很容易知道正确的命令。不幸的是我对 Pandas 还很陌生。即使只是提示在文档中查找位置也会很棒!

Time                   some value

2010-08-31 12:30:00    33.910
2010-08-31 12:40:00    33.250
2010-08-31 12:50:00    30.500
2010-08-31 13:00:00    27.065
2010-08-31 13:10:00    25.610
...

2013-06-07 02:10:00    16.970
2013-06-07 02:20:00    16.955
2013-06-07 02:30:00    17.000
2013-06-07 02:40:00    17.015
2013-06-07 02:50:00    16.910

Answer 1

您可以尝试groupby按小时和分钟和transform 意思:

print df
                  Time  some value
0  2010-08-31 12:30:00      33.910
1  2010-08-31 12:40:00      33.250
2  2010-08-31 12:50:00      30.500
3  2010-08-31 13:00:00      27.065
4  2010-08-31 13:10:00      25.610
5  2013-06-07 02:10:00      16.970
6  2013-06-07 02:20:00      16.955
7  2013-06-07 02:30:00      17.000
8  2013-06-07 02:40:00      17.015
9  2013-06-07 02:50:00      16.910

#convert column time to datetime
df['Time'] = pd.to_datetime(df['Time'])
#set index from column Time
df = df.set_index('Time')
print df
                     some value
Time                           
2010-08-31 12:30:00      33.910
2010-08-31 12:40:00      33.250
2010-08-31 12:50:00      30.500
2010-08-31 13:00:00      27.065
2010-08-31 13:10:00      25.610
2013-06-07 02:10:00      16.970
2013-06-07 02:20:00      16.955
2013-06-07 02:30:00      17.000
2013-06-07 02:40:00      17.015
2013-06-07 02:50:00      16.910

print df.groupby([df.index.hour, df.index.minute])['some value'].transform('mean')
Time
2010-08-31 12:30:00    33.910
2010-08-31 12:40:00    33.250
2010-08-31 12:50:00    30.500
2010-08-31 13:00:00    27.065
2010-08-31 13:10:00    25.610
2013-06-07 02:10:00    16.970
2013-06-07 02:20:00    16.955
2013-06-07 02:30:00    17.000
2013-06-07 02:40:00    17.015
2013-06-07 02:50:00    16.910
dtype: float64

下一个解决方案不将 index 设置为 Datetimeindex，使用 dt.hour和 dt.minute并创建新列newCol:

print df
                  Time  some value
0  2010-08-31 12:30:00      33.910
1  2010-08-31 12:40:00      33.250
2  2010-08-31 12:50:00      30.500
3  2010-08-31 13:00:00      27.065
4  2010-08-31 13:10:00      25.610
5  2013-06-07 02:10:00      16.970
6  2013-06-07 02:20:00      16.955
7  2013-06-07 02:30:00      17.000
8  2013-06-07 02:40:00      17.015
9  2013-06-07 02:50:00      16.910

#convert column time to datetime
df['Time'] = pd.to_datetime(df['Time'])
print df
                 Time  some value
0 2010-08-31 12:30:00      33.910
1 2010-08-31 12:40:00      33.250
2 2010-08-31 12:50:00      30.500
3 2010-08-31 13:00:00      27.065
4 2010-08-31 13:10:00      25.610
5 2013-06-07 02:10:00      16.970
6 2013-06-07 02:20:00      16.955
7 2013-06-07 02:30:00      17.000
8 2013-06-07 02:40:00      17.015
9 2013-06-07 02:50:00      16.910

df['newCol'] = df.groupby([df['Time'].dt.hour, df['Time'].dt.minute])['some value']
                 .transform('mean')
print df
                 Time  some value  newCol
0 2010-08-31 12:30:00      33.910  33.910
1 2010-08-31 12:40:00      33.250  33.250
2 2010-08-31 12:50:00      30.500  30.500
3 2010-08-31 13:00:00      27.065  27.065
4 2010-08-31 13:10:00      25.610  25.610
5 2013-06-07 02:10:00      16.970  16.970
6 2013-06-07 02:20:00      16.955  16.955
7 2013-06-07 02:30:00      17.000  17.000
8 2013-06-07 02:40:00      17.015  17.015
9 2013-06-07 02:50:00      16.910  16.910