計算時間序列數據的統計指標

Question

我有一個看起來像這樣的 dataframe：

        ID          DATE       QTD
0       71896517    2020-07-25  1
1       71896517    2020-09-14  2
2       72837949    2020-09-21  1
3       72848188    2020-11-03  1
4       73307986    2020-11-04  1
5       72848188    2020-11-16  1
6       71896517    2020-11-22  1
7       73307986    2020-11-25  1
8       73307986    2021-01-04  1
9       73307986    2021-02-04  1
10      72848188    2021-02-07  1
11      72837949    2021-02-11  1

我想在 dataframe 中添加一列作為QTD列的月平均值。 該指標應針對每個ID單獨計算（因此每個ID都有自己的值），並且應隨着 dataframe 在DATE （即年-月-日）中的增長而遞增。 如果ID在某個月份沒有活動，我希望該月仍包含在計算的指標中。

我希望結束 dataframe 看起來像這樣：

        ID          DATE       QTD    MEAN
0       71896517    2020-07-25  1     1.0
1       71896517    2020-09-14  2     1.0
2       72837949    2020-09-21  1     1.0
3       72848188    2020-11-03  1     1.0
4       73307986    2020-11-04  1     1.0
5       72848188    2020-11-16  1     2.0
6       71896517    2020-11-22  1     0.8
7       73307986    2020-11-25  1     2.0
8       73307986    2021-01-04  1     1.0
9       73307986    2021-02-04  1     1.0
10      72848188    2021-02-07  1     0.75
11      72837949    2021-02-11  1     0.33

我該如何 go 來實現這個？

Answer 1

通常有一個迭代的解決方案。

import pandas as pd
import numpy as np

def monthdelta(a,b):
    a1,a2,a3 = (int(k) for k in a.split('-'))
    b1,b2,b3 = (int(k) for k in b.split('-'))
    return (a1*12+a2) - (b1*12+b2)

data = [
[ 71896517, "2020-07-25", 1 ],
[ 71896517, "2020-09-14", 2 ],
[ 72837949, "2020-09-21", 1 ],
[ 72848188, "2020-11-03", 1 ],
[ 73307986, "2020-11-04", 1 ],
[ 72848188, "2020-11-16", 1 ],
[ 71896517, "2020-11-22", 1 ],
[ 73307986, "2020-11-25", 1 ],
[ 73307986, "2021-01-04", 1 ],
[ 73307986, "2021-02-04", 1 ],
[ 72848188, "2021-02-07", 1 ],
[ 72837949, "2021-02-11", 1 ],
]

df = pd.DataFrame( data, columns=["ID", "DATE", "QTD"] )

startdate = {}
sums = {}
sumsqs = {}
num = {}
stdev = []
means = []

for row in df.T.iteritems():
    id = row[1]['ID']
    if id not in startdate:
        num[id] = 1
        startdate[id] = row[1]['DATE']
        sums[id] = row[1]['QTD']
        sumsqs[id] = row[1]['QTD'] * row[1]['QTD']
        means.append( row[1]['QTD'] )
        stdev.append( 0 )
    else:
        num[id] += 1
        sums[id] += row[1]['QTD']
        sumsqs[id] += row[1]['QTD'] * row[1]['QTD']
        delta = monthdelta(row[1]['DATE'],startdate[id]) + 1
        means.append( sums[id] / delta )
        if delta == 1:
            stdev.append( 0 )
        else:
            stdev.append( np.sqrt((delta*sumsqs[id] - sums[id]*sums[id])/delta))

print( means )
df['MEAN'] = pd.Series(means)
print( stdev )
df['STDEV'] = pd.Series(stdev)
print( df )

Output：

[1, 1.0, 1, 1, 1, 2.0, 0.8, 2.0, 1.0, 1.0, 0.75, 0.3333333333333333]
[0, 1.4142135623730951, 0, 0, 0, 0, 1.6733200530681511, 0, 0.0, 0.0, 0.8660254037844386, 1.1547005383792515]
          ID        DATE  QTD      MEAN     STDEV
0   71896517  2020-07-25    1  1.000000  0.000000
1   71896517  2020-09-14    2  1.000000  1.414214
2   72837949  2020-09-21    1  1.000000  0.000000
3   72848188  2020-11-03    1  1.000000  0.000000
4   73307986  2020-11-04    1  1.000000  0.000000
5   72848188  2020-11-16    1  2.000000  0.000000
6   71896517  2020-11-22    1  0.800000  1.673320
7   73307986  2020-11-25    1  2.000000  0.000000
8   73307986  2021-01-04    1  1.000000  0.000000
9   73307986  2021-02-04    1  1.000000  0.000000
10  72848188  2021-02-07    1  0.750000  0.866025
11  72837949  2021-02-11    1  0.333333  1.154701