问题： 我需要平均给定年份中每5周一次的商品和商店销售额。平均值也必须能够“循环”多周的1,2,51,52周。
例如;第1周将是平均周数：51,52,1,2和3，产品72243000016，在商店10103中。
>>> df
weekNumber productNumber storeNumber Sales
0 47.0 72243000016 10103 93.80
1 47.0 72243000016 10148 97.43
2 47.0 72243000016 10153 114.01
3 47.0 72243000016 10216 154.75
4 47.0 72243000016 10243 55.74
5 47.0 72243000016 10260 52.74
6 47.0 72243000016 10266 104.38
7 47.0 72243000016 10275 80.06
8 47.0 72243000016 10327 40.11
9 47.0 72243000016 10375 57.32
10 47.0 72243000016 10402 25.58
11 47.0 72243000016 10407 51.32
12 47.0 72243000016 10412 13.58
13 47.0 72243000016 10436 86.22
14 47.0 72243000016 10537 32.53
15 47.0 72243000016 10588 41.37
16 47.0 72243010016 10103 76.27
17 47.0 72243010016 10148 61.27
18 47.0 72243010016 10153 96.64
19 47.0 72243010016 10216 75.48
20 47.0 72243010016 10243 39.95
21 47.0 72243010016 10260 47.53
22 47.0 72243010016 10266 37.74
23 47.0 72243010016 10275 56.69
24 47.0 72243010016 10327 17.37
25 47.0 72243010016 10375 22.58
26 47.0 72243010016 10402 29.53
27 47.0 72243010016 10436 46.11
28 47.0 72243010016 10537 27.16
29 47.0 72243010016 10588 33.16
... ... ... ... ...
118039 5.0 85005700315 10275 30.72
118040 5.0 85005700315 10402 11.97
118041 5.0 85005700315 10436 35.51
118042 5.0 85005700315 10412 19.95
118043 5.0 85005700315 10148 67.43
118044 5.0 85005700315 10260 47.48
118045 5.0 85005700315 10103 67.43
118046 5.0 85005700315 10327 7.98
118047 5.0 85005700315 10216 83.79
118048 5.0 85005700319 10637 19.95
118049 5.0 85005700319 10266 23.94
118050 5.0 85005700319 10537 19.95
118051 5.0 85005700319 10243 39.90
118052 5.0 85005700319 10275 35.51
118053 5.0 85005700319 10402 15.96
118054 5.0 85005700319 10436 35.51
118055 5.0 85005700319 10148 19.95
118056 5.0 85005700319 10103 119.60
118057 5.0 85005700319 10327 3.99
118058 5.0 85005700319 10216 151.42
118060 5.0 85005700324 10260 42.99
118061 5.0 85005700340 10637 63.84
118062 5.0 85005700340 10266 47.88
118063 5.0 85005700340 10537 7.98
118064 5.0 85005700340 10275 90.97
118065 5.0 85005700340 10402 23.84
118066 5.0 85005700340 10436 103.34
118067 5.0 85005700340 10148 43.09
118068 5.0 85005700340 10103 147.03
118069 5.0 85005700340 10327 7.88
这是有效的，但我相信有一种更有效的方法。
>>> def querythis(weekNumber, productNumber, storeNumber):
a=[-2,-1,0,1,2]
weekNumber=[(52+weekNumber+x)%52 for x in a]
weekNumber=[52 if x==0 else x for x in weekNumber]
WeekSum=0
count=0
for i in weekNumber:
YoYWeek=df[(df['weekNumber']==i)\
&(df['productNumber']==productNumber)\
&(df['Store Number']==storeNumber)][[3]].mean()
if YoYWeek.any():
WeekSum= WeekSum + YoYWeek
count=count+1
return WeekSum/count
>>> df['Rolling Average']=df.apply(lambda row:querythis(row['week_num'],row['productNumber'],row['Store Number']), axis=1)