Question

我有一个数据框，如下所示。我想使用“ part1”列作为基准，将数据分类为3个部分（每个部分具有相同的数字数据集）并计算每个组的part2平均值。如row0和row1作为groupB，平均值为（0.67 +（-0.03））/ 2。

import pandas as pd
df = pd.DataFrame({
    "date":["20130101","20130101","20130103","20130103","20130105","20130105"],
    "part1":[0.5,0.7,1.3,1.5,0.1,0.3],
    "part2":[0.67,-0.03,1.95,-3.25,-0.3,0.6]
})
    date    part1   part2   output
0   20130101    0.5 0.67    0.32
1   20130101    0.7 -0.03   0.32
2   20130103    1.3 1.95    -0.65
3   20130103    1.5 -3.25   -0.65
4   20130105    0.1 -0.3    0.15
5   20130105    0.3 0.6 0.15

Answer 1

如果您的数据存储在pandas中，则可以使用df进行操作：

def foo(x,n=3):
    df = x.copy()
    bins = np.quantile(df['part1'],np.linspace(0,1,n+1))
    df['tmp'] = 0
    for i in range(n):
        idx = (df['part1'] > bins[i]) & (df['part1'] <= bins[i+1])
        df['tmp'][idx] = i
    return df.groupby('tmp').agg({'part2':'mean'})

foo(df)

将与分位数分开，因此保证，每个组将具有相同数量的元素。按tmp分组将得到这些分组，并计算每个分组的part2的平均值：

     part2
tmp       
0     0.15
1     0.32
2    -0.65

Answer 2

如果要计算每天的平均值，可以使用groupby，如下所示：

import pandas as pd
df = pd.DataFrame({
    "date":["20130101","20130101","20130103","20130103","20130105","20130105"],
    "part1":[0.5,0.7,1.3,1.5,0.1,0.3],
    "part2":[0.67,-0.03,1.95,-3.25,-0.3,0.6]
})
df.groupby("date").mean().reset_index()

结果：

       date  part1  part2
0  20130101    0.6   0.32
1  20130103    1.4  -0.65
2  20130105    0.2   0.15

Answer 3

您可以为熊猫的by方法的groupby参数传递函数。

from functools import partial
import pandas as pd

df = pd.DataFrame({
    "date":["20130101","20130101","20130103","20130103","20130105","20130105"],
    "part1":[0.5,0.7,1.3,1.5,0.1,0.3],
    "part2":[0.67,-0.03,1.95,-3.25,-0.3,0.6]
})


def grouper(df, val):
    foo = df.iloc[val]['part1']
    if foo > 0.0 and foo < 0.4:
         return 0
    elif foo > 0.3 and foo < 1.0:
         return 1
    elif foo > 1.0:
         return 2

grouped = df['part2'].groupby(by=partial(grouper, df)).mean()

这导致

1    0.15
2    0.32
3   -0.65
Name: part2, dtype: float64

python如何使用groupby对数据进行分类并计算其他列的均值

3 个答案: