在《Python Pandas窗口函数》一节，我们重点介绍了窗口函数。我们知道，窗口函数可以与聚合函数一起使用，聚合函数指的是对一组数据求总和、最大值、最小值以及平均值的操作，本节重点讲解聚合函数的应用。
应用聚合函数
首先让我们创建一个 DataFrame 对象，然后对聚合函数进行应用。
- import pandas as pd
import numpy as np
df = pd.DataFrame(np.random.randn(5, 4),index = pd.date_range('12/14/2020', periods=5),columns = ['A', 'B', 'C', 'D'])
print (df)
#窗口大小为3，min_periods 最小观测值为1
r = df.rolling(window=3,min_periods=1)
print(r)
输出结果：
A B C D
2020-12-14 0.941621 1.205489 0.473771 -0.348169
2020-12-15 -0.276954 0.076387 0.104194 1.537357
2020-12-16 0.582515 0.481999 -0.652332 -1.893678
2020-12-17 -0.286432 0.923514 0.285255 -0.739378
2020-12-18 2.063422 -0.465873 -0.946809 1.590234 Rolling [window=3,min_periods=1,center=False,axis=0]
1) 对整体聚合
您可以把一个聚合函数传递给 DataFrame，示例如下：
- import pandas as pd
import numpy as np
df = pd.DataFrame(np.random.randn(5, 4),index = pd.date_range('12/14/2020', periods=5),columns = ['A', 'B', 'C', 'D'])
print (df)
#窗口大小为3，min_periods 最小观测值为1
r = df.rolling(window=3,min_periods=1)
#使用 aggregate()聚合操作
print(r.aggregate(np.sum))
输出结果：
A B C D
2020-12-14 0.133713 0.746781 0.499385 0.589799
2020-12-15 -0.777572 0.531269 0.600577 -0.393623
2020-12-16 0.408115 -0.874079 0.584320 0.507580
2020-12-17 -1.033055 -1.185399 -0.546567 2.094643
2020-12-18 0.469394 -1.110549 -0.856245 0.260827 A B C D
2020-12-14 0.133713 0.746781 0.499385 0.589799
2020-12-15 -0.643859 1.278050 1.099962 0.196176
2020-12-16 -0.235744 0.403971 1.684281 0.703756
2020-12-17 -1.402513 -1.528209 0.638330 2.208601
2020-12-18 -0.155546 -3.170027 -0.818492 2.863051
2) 对任意某一列聚合
- import pandas as pd
import numpy as np
df = pd.DataFrame(np.random.randn(5, 4),index = pd.date_range('12/14/2020', periods=5),columns = ['A', 'B', 'C', 'D'])
#窗口大小为3，min_periods 最小观测值为1
r = df.rolling(window=3,min_periods=1)
#对 A 列聚合
print(r['A'].aggregate(np.sum))
输出结果：
2020-12-14 1.051501
2020-12-15 1.354574
2020-12-16 0.896335
2020-12-17 0.508470
2020-12-18 2.333732
Freq: D, Name: A, dtype: float64
3) 对多列数据聚合
- import pandas as pd
import numpy as np
df = pd.DataFrame(np.random.randn(5, 4),index = pd.date_range('12/14/2020', periods=5),columns = ['A', 'B', 'C', 'D'])
#窗口大小为3，min_periods 最小观测值为1
r = df.rolling(window=3,min_periods=1)
#对 A/B 两列聚合
print(r['A','B'].aggregate(np.sum))
输出结果：
A B
2020-12-14 0.639867 -0.229990
2020-12-15 0.352028 0.257918
2020-12-16 0.637845 2.643628
2020-12-17 0.432715 2.428604
2020-12-18 -1.575766 0.969600
4) 对单列应用多个函数
- import pandas as pd
import numpy as np
df = pd.DataFrame(np.random.randn(5, 4),index = pd.date_range('12/14/2020', periods=5),columns = ['A', 'B', 'C', 'D'])
#窗口大小为3，min_periods 最小观测值为1
r = df.rolling(window=3,min_periods=1)
#对 A/B 两列聚合
print(r['A','B'].aggregate([np.sum,np.mean]))
输出结果：
sum mean
2020-12-14 -0.469643 -0.469643
2020-12-15 -0.626856 -0.313428
2020-12-16 -1.820226 -0.606742
2020-12-17 -2.007323 -0.669108
2020-12-18 -0.595736 -0.198579
5) 对不同列应用多个函数
- import pandas as pd
import numpy as np
df = pd.DataFrame(np.random.randn(5, 4),
index = pd.date_range('12/11/2020', periods=5),
columns = ['A', 'B', 'C', 'D'])
r = df.rolling(window=3,min_periods=1)
print( r['A','B'].aggregate([np.sum,np.mean]))
输出结果：
A B
sum mean sum mean
2020-12-14 -1.428882 -1.428882 -0.417241 -0.417241
2020-12-15 -1.315151 -0.657576 -1.580616 -0.790308
2020-12-16 -2.093907 -0.697969 -2.260181 -0.753394
2020-12-17 -1.324490 -0.441497 -1.578467 -0.526156
2020-12-18 -2.400948 -0.800316 -0.452740 -0.150913
6) 对不同列应用不同函数
- import pandas as pd
import numpy as np
df = pd.DataFrame(np.random.randn(3, 4),
index = pd.date_range('12/14/2020', periods=3),
columns = ['A', 'B', 'C', 'D'])
r = df.rolling(window=3,min_periods=1)
print(r.aggregate({'A': np.sum,'B': np.mean}))
输出结果：
A B
2020-12-14 0.503535 -1.301423
2020-12-15 0.170056 -0.550289
2020-12-16 -0.086081 -0.140532
