1.init a dataframe
import pandas as pd
import numpy as np
df = pd.DataFrame({'key1':['a', 'a', 'b', 'b', 'a','a','a','a'],
'key2':['one', 'two', 'one', 'two', 'one','two','two','two'],
'data1':np.random.randn(8),
'data2':np.random.randn(8)})
df.head(10)
2.data1按key1，key2分组后的平均值
df_group = df.groupby(["key1","key2"])["data1"].mean()
print(df_group)
result:
key1 key2
a one 1.197223
two -0.224934
b one 1.779484
two 1.193350
Name: data1, dtype: float64
3.group 操作后还恢复到正常dataframe索引
df_reset = df_group.reset_index()
df_reset.head()
result:
key1 key2 data1
0 a one 1.197223
1 a two -0.224934
2 b one 1.779484
3 b two 1.193350
4.同一组数据做多类型数据统计
df_group_m = df.groupby(["key1","key2"])["data1"].agg(["mean","max"])
df_reset_m = df_group_m.reset_index()
print(df_group_m)
df_reset_m.head()
result:
mean max
key1 key2
a one 1.197223 2.319615
two -0.224934 -0.173460
b one 1.779484 1.779484
two 1.193350 1.357037
key1 key2 mean max
0 a one 1.197223 2.319615
1 a two -0.224934 -0.173460
2 b one 1.779484 1.779484
3 b two 1.193350 1.357037
5.自定义统计函数
系统提供了丰富的统计函数，比如：最大值、最小值、count、求和、平均值等常见的统计属性，但是有时候仍然不能满足我们的需求，需要自己给grouby写统计函数（这里要特别小心，在数据量大时，任何的时间消耗都会被放大，统计要尽可能简单）。
比如要获取大于平均数的数据的中位数附近的数据。
def median_m(arrs):
length = len(arrs) - 1
idx = int(0.75*length)
return arrs.iat[idx]
df.sort_values(by=["data1"],inplace=True,ascending=True)
df_group_d = df.groupby(["key1","key2"])["data1"].agg(["mean",median_m])
df_reset_d = df_group_d.reset_index()
print(df_group_d)
df_reset_d.head()
result:
mean median_m
key1 key2
a one 1.223088 0.146005
two -1.033850 -0.334123
b one -0.194909 -0.194909
two -1.798123 -1.798123
key1 key2 mean median_m
0 a one 1.223088 0.146005
1 a two -1.033850 -0.334123
2 b one -0.194909 -0.194909
3 b two -1.798123 -1.798123
