python - 如何像它的代码一样显示 pandas 滚动窗口？

Question

是的，我在 this question 的问题中修改了我的 pandas 数据框示例:但是这次的问题不一样了

import pandas as pd
import numpy as np

d=({'A':['2020-01-01','2020-01-02','2020-01-03','2020-01-04','2020-01-05','2020-01-06','2020-01-07','2020-01-08','2020-01-09','2020-01-10'],
'B':[100,110,120,175,164,169,155,153,156,200]})
df=pd.DataFrame(d)

            A    B
0  2020-01-01  100
1  2020-01-02  110
2  2020-01-03  120
3  2020-01-04  175
4  2020-01-05  164
5  2020-01-06  169
6  2020-01-07  155
7  2020-01-08  153
8  2020-01-09  156
9  2020-01-10  200

我知道 pandas 滚动窗口代码是这样的(对于我的情况):df['sums']=df['B'].rolling(window=3, center=False).sum()

            A    B   sums
0  2020-01-01  100    NaN #Here's the question:
1  2020-01-02  110    NaN #Why the rolling windows NaN is just 2, but the rolling window code is 3
2  2020-01-03  120  330.0
3  2020-01-04  175  405.0
4  2020-01-05  164  459.0
5  2020-01-06  169  508.0
6  2020-01-07  155  488.0
7  2020-01-08  153  477.0
8  2020-01-09  156  464.0
9  2020-01-10  200  509.0
Name: sum, dtype: float64

我想把它变成这样:

0  2020-01-01  100    NaN # To display rolling window
1  2020-01-02  110    NaN # Like the code
2  2020-01-03  120    NaN # Not 2 but 3 NaNs
3  2020-01-04  175  330.0
4  2020-01-05  164  405.0
5  2020-01-06  169  459.0
6  2020-01-07  155  508.0
7  2020-01-08  153  488.0
8  2020-01-09  156  477.0
9  2020-01-10  200  464.0
10 2020-01-11  NaN  509.0

有办法吗？我尝试添加 NaN=np.nan 它变成这样:

1  2020-01-02  110    NaN
2  2020-01-03  120    NaN 
3  2020-01-04  175  330.0
.
.
9  2020-01-10  200  464.0
10 2020-01-11  NaN  509.0
0  2020-01-01  100    NaN

有什么解决办法吗？

Answer 1

系列开头有两个 NaN，因为此时您没有足够的数据进行滚动(1 和 2 而不是 3)。从第 3 行开始。

如果您希望 3 以 NaN 开头，请使用 shift 链接您的滚动代码:

df['sums']=df['A'].rolling(window=3, center=False).sum().shift()

输出:

     A   sums
0  100    NaN
1  110    NaN
2  120    NaN
3  175  330.0
4  164  405.0
5  169  459.0
6  155  508.0
7  153  488.0
8  156  477.0
9  200  464.0