Question

我有这样的数据框，

import pandas as pd

d = {'ID':["A","A","A","A","A","A","A","A","A","A","A","A"],
     'date':["2017-01-01","2017-01-01","2017-01-01","2017-01-02","2017-01-02","2017-01-02","2017-01-03","2017-01-03",
         "2017-01-03","2017-01-04","2017-01-04","2017-01-04"],
     'time':["00:00","06:00","12:00","00:00","06:00","12:00","00:00","06:00","12:00","00:00","06:00","12:00"],
     'value':[23,100,330,57,122,477,46,99,469,37,118,499]}

df = pd.DataFrame(data=d)

df['date'] =  pd.to_datetime(df['date'])
print(df)

  ID        date   time   value
0   A  2017-01-01  00:00     23
1   A  2017-01-01  06:00    100
2   A  2017-01-01  12:00    330
3   A  2017-01-02  00:00     57
4   A  2017-01-02  06:00    122
5   A  2017-01-02  12:00    477
6   A  2017-01-03  00:00     46
7   A  2017-01-03  06:00     99
8   A  2017-01-03  12:00    469
9   A  2017-01-04  00:00     37
10  A  2017-01-04  06:00    118
11  A  2017-01-04  12:00    499

我想生成一个包含基于时间顺序的历史数据的新列。最终的数据框架就像这样，

      ID        date   time   value   avg
    0   A  2017-01-01  00:00     23    23
    1   A  2017-01-01  06:00    100   100
    2   A  2017-01-01  12:00    330   330
    3   A  2017-01-02  00:00     57    23
    4   A  2017-01-02  06:00    122   100
    5   A  2017-01-02  12:00    477   330
    6   A  2017-01-03  00:00     46    40     # (23+57)/2 = 40
    7   A  2017-01-03  06:00     99   111     # (100+122)/2 = 111
    8   A  2017-01-03  12:00    469  403.5    # (330+477)/2 = 403.5
    9   A  2017-01-04  00:00     37    42     # (23+57+46)/3 = 42
    10  A  2017-01-04  06:00    118   107     # (100+122+99)/3 = 107
    11  A  2017-01-04  12:00    499  425.3    # (330+477+469)/3 = 425.333

新列平均值计算同一历史时间点数据的平均值。所以，前两天是相同的 - 只需复制第一天的数据。然后第三天将是前两天的平均值，依此类推。

这只是一个示例数据集。我希望有人可以有一般功能来解决这个问题。谢谢！

Answer 1

IIUC，让我们试试这个：

df.set_index(['ID','time','date'])['value']\
  .unstack([0,1])\
  .rolling(len(df),min_periods=1)\
  .mean().shift(1).bfill()\
  .unstack().rename('avg')\
  .to_frame()\
  .join(df.set_index(['ID','time','date']))\
  .reset_index().sort_values(['ID','date','time'])

输出：

    ID  time    date    avg value
0   A   00:00   2017-01-01  23.000000   23
4   A   06:00   2017-01-01  100.000000  100
8   A   12:00   2017-01-01  330.000000  330
1   A   00:00   2017-01-02  23.000000   57
5   A   06:00   2017-01-02  100.000000  122
9   A   12:00   2017-01-02  330.000000  477
2   A   00:00   2017-01-03  40.000000   46
6   A   06:00   2017-01-03  111.000000  99
10  A   12:00   2017-01-03  403.500000  469
3   A   00:00   2017-01-04  42.000000   37
7   A   06:00   2017-01-04  107.000000  118
11  A   12:00   2017-01-04  425.333333  499

如何按python数据帧中的历史时间序列值聚合数据？

1 个答案: