我试图在Pandas DataFrame中添加两个列 - 一个代表下面几周的值，另一个代表下面4周的总和。
现有DataFrame的示例如下所示。下面的DataFrame只是整个DataFrame的一个简短片段，它跨越了多年。下面的DataFrame使用以下函数派生：df = df.groupby([pd.Grouper(key='date', freq='W'), pd.Grouper('company_name').agg({'returns': 'sum'})
date company_name returns
2014-12-07 Amazon -0.5
2014-12-14 Amazon -0.1
2014-12-21 Amazon 0.5
2014-12-28 Amazon 0.3
2015-01-04 Amazon 0.1
2014-12-07 Facebook 0.5
2014-12-14 Facebook 0.5
2014-12-21 Facebook 0.5
2014-12-28 Facebook -0.5
2015-01-04 Facebook -0.5
2014-12-07 Google 0.1
2014-12-14 Google 0.1
2014-12-21 Google 0.1
2014-12-28 Google 0.1
2015-01-04 Google 0.1
2014-12-07 Intel 0.2
2014-12-14 Intel 0.2
2014-12-21 Intel 0.2
2014-12-28 Intel 0.2
2015-01-04 Intel 0.2期望的输出将返回下周的值，以及从“日期”列开始的接下来的4周的总和。下面可以看到所需输出的一个例子。
date company_name returns next_week_return next_month_return
2014-12-07 Amazon -0.5 -0.5 0.8
2014-12-14 Amazon -0.1 0.5 0.8
2014-12-21 Amazon 0.5 0.3 0.8
2014-12-28 Amazon 0.3 0.1 0.8
2015-01-04 Amazon 0.1 0.1 ...
2014-12-07 Facebook 0.5 0.5 0.0
2014-12-14 Facebook 0.5 0.5 0.0
2014-12-21 Facebook 0.5 -0.5 0.0
2014-12-28 Facebook -0.5 -0.5 0.0
2015-01-04 Facebook -0.5 0.1 ...
2014-12-07 Google 0.1 0.1 0.4
2014-12-14 Google 0.1 0.1 0.4
2014-12-21 Google 0.1 0.1 0.4
2014-12-28 Google 0.1 0.1 0.4
2015-01-04 Google 0.1 0.1 ...
2014-12-07 Intel 0.2 0.2 0.8
2014-12-14 Intel 0.2 0.2 0.8
2014-12-21 Intel 0.2 0.2 0.8
2014-12-28 Intel 0.2 0.2 0.8
2015-01-04 Intel 0.2 0.2可以在下面看到原始CSV的片段。
date CompanyName return
07/12/2014 8x8 Inc -0.0038835
14/12/2014 8x8 Inc 0.036923354
21/12/2014 8x8 Inc 0.108854405
28/12/2014 8x8 Inc 0.042793145
04/01/2015 8x8 Inc -0.027219971
11/01/2015 8x8 Inc -0.038249882
18/01/2015 8x8 Inc 0.045946457
25/01/2015 8x8 Inc -0.107796707
01/02/2015 8x8 Inc -0.056725981
08/02/2015 8x8 Inc 0.024344572
15/02/2015 8x8 Inc 0.00756624
22/02/2015 8x8 Inc -0.04365263
01/03/2015 8x8 Inc -0.02794593
08/03/2015 8x8 Inc -0.039922714
15/03/2015 8x8 Inc 0.020848566
22/03/2015 8x8 Inc 0.116712617
29/03/2015 8x8 Inc 0.028952565
05/04/2015 8x8 Inc 0.053253322
12/04/2015 8x8 Inc -0.006787356
19/04/2015 8x8 Inc -0.00912207
26/04/2015 8x8 Inc 0.013652089
03/05/2015 8x8 Inc -0.021702736
10/05/2015 8x8 Inc -0.021004273
17/05/2015 8x8 Inc 0.012888286
24/05/2015 8x8 Inc -0.021177262
31/05/2015 8x8 Inc -0.027630051
07/12/2014 AB SA -1.015859196
14/12/2014 AB SA -0.01810143
21/12/2014 AB SA -0.073869849
28/12/2014 AB SA 0.000666445
04/01/2015 AB SA 0.051293294
11/01/2015 AB SA 0.004735605
18/01/2015 AB SA 0.014073727
25/01/2015 AB SA 0.097002705
01/02/2015 AB SA 0.00337648
08/02/2015 AB SA 0.018093743
15/02/2015 AB SA 0.019667392
22/02/2015 AB SA 0.024844339
01/03/2015 AB SA 0.015707129
08/03/2015 AB SA 0.109611209
15/03/2015 AB SA -0.039164849
22/03/2015 AB SA -0.002909093
29/03/2015 AB SA 0.007256926
05/04/2015 AB SA -0.025385791
12/04/2015 AB SA 0.019584469
19/04/2015 AB SA -0.01342302
26/04/2015 AB SA 0.073405725
03/05/2015 AB SA -0.018666287
10/05/2015 AB SA 0.019350984
17/05/2015 AB SA -0.030814439
24/05/2015 AB SA 0.027386256
31/05/2015 AB SA -0.033285978
07/12/2014 ACCO Brands Corp 0.432332004
14/12/2014 ACCO Brands Corp -0.064822249
21/12/2014 ACCO Brands Corp 0.010163837
28/12/2014 ACCO Brands Corp 0.022223137
04/01/2015 ACCO Brands Corp -0.034659702
11/01/2015 ACCO Brands Corp -0.026514522
18/01/2015 ACCO Brands Corp -0.018868484
25/01/2015 ACCO Brands Corp 0.013010237
01/02/2015 ACCO Brands Corp -0.071850737
08/02/2015 ACCO Brands Corp 0.00126183
15/02/2015 ACCO Brands Corp -0.016000601
22/02/2015 ACCO Brands Corp -0.01420295
01/03/2015 ACCO Brands Corp -0.010457612
08/03/2015 ACCO Brands Corp -0.006591982
15/03/2015 ACCO Brands Corp -0.008257798
22/03/2015 ACCO Brands Corp 0.039272062
29/03/2015 ACCO Brands Corp 0.035312622
05/04/2015 ACCO Brands Corp 0.012315427
12/04/2015 ACCO Brands Corp 0.037241541
19/04/2015 ACCO Brands Corp -0.025075941
26/04/2015 ACCO Brands Corp -0.010535083
03/05/2015 ACCO Brands Corp -0.044016885
10/05/2015 ACCO Brands Corp -0.013845407
17/05/2015 ACCO Brands Corp 0.005056901
24/05/2015 ACCO Brands Corp -0.024251348
31/05/2015 ACCO Brands Corp -0.051701374
07/12/2014 Acer Inc 3.829777429
07/12/2014 Acer Inc -3.46435286
14/12/2014 Acer Inc 0.042160811
14/12/2014 Acer Inc 0.021342273
21/12/2014 Acer Inc -0.056618894
21/12/2014 Acer Inc -0.046304568
28/12/2014 Acer Inc 0.033415997
28/12/2014 Acer Inc 0.062759689
04/01/2015 Acer Inc 0.002344667
04/01/2015 Acer Inc -0.004460974
11/01/2015 Acer Inc 0.082988363
11/01/2015 Acer Inc 0.093933758
18/01/2015 Acer Inc -0.033983853
18/01/2015 Acer Inc -0.042689409
25/01/2015 Acer Inc 0.017136282
25/01/2015 Acer Inc -0.012539349
01/02/2015 Acer Inc 0.002424244
01/02/2015 Acer Inc 0.010980502
08/02/2015 Acer Inc -0.014634408
08/02/2015 Acer Inc -0.015723594
15/02/2015 Acer Inc -0.014851758
15/02/2015 Acer Inc 0.025040432
22/02/2015 Acer Inc 0
22/02/2015 Acer Inc 0.022919261
01/03/2015 Acer Inc 0.024631787
01/03/2015 Acer Inc -0.007581537
08/03/2015 Acer Inc 0.05445132
08/03/2015 Acer Inc 0.027028672
15/03/2015 Acer Inc -0.023311079
15/03/2015 Acer Inc -0.022472856
22/03/2015 Acer Inc -0.002361276
22/03/2015 Acer Inc 0
29/03/2015 Acer Inc -0.021506205
29/03/2015 Acer Inc 0.012048339
05/04/2015 Acer Inc -0.021978907
05/04/2015 Acer Inc -0.028109292
12/04/2015 Acer Inc -0.004950505
12/04/2015 Acer Inc 0.02756683
19/04/2015 Acer Inc -0.007472015
19/04/2015 Acer Inc 0.003016594
26/04/2015 Acer Inc 0.009950331
26/04/2015 Acer Inc 0.006006024
03/05/2015 Acer Inc -0.004962789
03/05/2015 Acer Inc 0.002989539
10/05/2015 Acer Inc -0.040614719
10/05/2015 Acer Inc -0.087282784
17/05/2015 Acer Inc -0.064193158
17/05/2015 Acer Inc -0.072605718
24/05/2015 Acer Inc 0.008253142
24/05/2015 Acer Inc -0.032031208
31/05/2015 Acer Inc 0.005464494
31/05/2015 Acer Inc 0.057961788从上面我希望每行添加两列 - 一个next_week_return，其中显示特定公司的下几周返回;另一个：next_month_return，这将是以下四周回归的总和。
任何人都可以提供的任何帮助将不胜感激。
从输入CSV的示例片段开始，一种解决方案是编写自定义函数以与df.apply()一起使用，该函数接受每个公司的子DataFrame，并且对于子DataFrame中的每个日期，计算指定的return的总和。前瞻天的数量。
以下代码假定df保存原始CSV中的示例数据。
# Convert string dates to pandas.Timestamp
df['date'] = pd.to_datetime(df['date'])
# Within each CompanyName, sort by date, because we'll
# set the date column as a DatetimeIndex and will
# index-slice it with pandas date offsets, and this
# requires a sorted index.
df.sort_values(['CompanyName', 'date'], inplace=True)
# Set a MultiIndex to ensure that the calculated
# columns returned by the custom function align correctly
df.set_index(['CompanyName', 'date'], inplace=True)
# Define a custom function to sum the values of `return` for
# each CompanyName sub-DataFrame. The defaults of 1 and 1+28
# capture the month (defined to be 28 days) immediately following
# each date, excluding the date itself. To get just the
# next week's values, use start=1, end=7.
def sum_return_over_next_i_to_j_days(df, first=1, last=1+28):
day = pd.offsets.Day(1)
df.reset_index(level=0, drop=True, inplace=True)
rets = [df.loc[today + first*day : today + last*day, 'return'].sum(min_count=1)
for today in df.index]
return pd.DataFrame(rets,
index=df.index,
columns=[f'sum_return_next_{first}-{last}_days'])
# Apply the above function to input CSV
df['next_week_return'] = df.groupby('CompanyName').apply(sum_return_over_next_i_to_j_days, 1, 7)
df['next_month_return'] = df.groupby('CompanyName').apply(sum_return_over_next_i_to_j_days, 1, 1+28)
df = df.reset_index()
# Print result
df.head(10)
CompanyName date return next_week_return next_month_return
0 8x8 Inc 2014-07-12 -0.003883 NaN NaN
1 8x8 Inc 2014-12-14 0.036923 0.108854 0.066976
2 8x8 Inc 2014-12-21 0.108854 0.042793 0.004068
3 8x8 Inc 2014-12-28 0.042793 -0.084672 -0.146522
4 8x8 Inc 2015-01-02 -0.056726 -0.027946 -0.089796
5 8x8 Inc 2015-01-03 -0.027946 NaN -0.061850
6 8x8 Inc 2015-01-18 0.045946 -0.107797 -0.100230
7 8x8 Inc 2015-01-25 -0.107797 NaN -0.036086
8 8x8 Inc 2015-02-15 0.007566 -0.043653 -0.044507
9 8x8 Inc 2015-02-22 -0.043653 NaN 0.115858
df.tail(10)
CompanyName date return next_week_return next_month_return
120 Acer Inc 2015-08-02 -0.014634 0.08148 0.081480
121 Acer Inc 2015-08-02 -0.015724 0.08148 0.081480
122 Acer Inc 2015-08-03 0.054451 NaN NaN
123 Acer Inc 2015-08-03 0.027029 NaN NaN
124 Acer Inc 2015-10-05 -0.040615 NaN 0.176922
125 Acer Inc 2015-10-05 -0.087283 NaN 0.176922
126 Acer Inc 2015-11-01 0.082988 NaN NaN
127 Acer Inc 2015-11-01 0.093934 NaN NaN
128 Acer Inc 2015-12-04 -0.004951 NaN NaN
129 Acer Inc 2015-12-04 0.027567 NaN NaN