Pandas+groupby用法讲解

import pandas as pd
import numpy as np

#导入数据
type_specified={"trip_id":"object","bikeid":"object","from_station_id":"object","to_station_id":"object",
               "hour":"object","weekday":"object"}
ridership=pd.read_csv(r"E:\chicago_bikes_data\txt&csv\ridership_2016.txt",
                      nrows=10000,dtype=type_specified,usecols=range(1,15))

ridership.head()

	trip_id	starttime	stoptime	bikeid	tripduration	from_station_id	from_station_name	to_station_id	to_station_name	usertype	gender	birthyear	hour	weekday
0	9080551	3/31/2016 23:53	4/1/2016 0:07	155	841	344	Ravenswood Ave & Lawrence Ave	458	Broadway & Thorndale Ave	Subscriber	Male	1986.0	23	4
1	9080550	3/31/2016 23:46	3/31/2016 23:57	4831	649	128	Damen Ave & Chicago Ave	213	Leavitt St & North Ave	Subscriber	Male	1980.0	23	4
2	9080549	3/31/2016 23:42	3/31/2016 23:46	4232	210	350	Ashland Ave & Chicago Ave	210	Ashland Ave & Division St	Subscriber	Male	1979.0	23	4
3	9080548	3/31/2016 23:37	3/31/2016 23:55	3464	1045	303	Broadway & Cornelia Ave	458	Broadway & Thorndale Ave	Subscriber	Male	1980.0	23	4
4	9080547	3/31/2016 23:33	3/31/2016 23:37	1750	202	334	Lake Shore Dr & Belmont Ave	329	Lake Shore Dr & Diversey Pkwy	Subscriber	Male	1969.0	23	4

#groupby可以通过传入需要分组的参数实现对数据的分组
ridership_user=ridership.groupby("usertype")
ridership_user_week=ridership.groupby(["usertype","weekday"])

#groupby之后的数据并不是DataFrame格式的数据，而是特殊的groupby类型，此时，可以通过size()方法返回分组后的记录数目统计结果，该结果
#是Series类型
ridership_user.size()

usertype
Customer      1023
Subscriber    8977
dtype: int64

ridership_user_week.size()

usertype    weekday
Customer    3           286
            4           737
Subscriber  3          2714
            4          6263
dtype: int64

#可以通过索引的方式实现对分组后各类别记录数的提取
ridership_user_week.size().loc["Customer","3"]

286

#通过groupby完成对数据的分组后，可以通过get_group方法来获取某一制定分组的结果

ridership_user.get_group("Customer").head()

	trip_id	starttime	stoptime	bikeid	tripduration	from_station_id	from_station_name	to_station_id	to_station_name	usertype	gender	birthyear	hour	weekday
23	9080526	3/31/2016 23:04	3/31/2016 23:32	1638	1648	505	Winchester Ave & Elston Ave	505	Winchester Ave & Elston Ave	Customer	NaN	NaN	23	4
65	9080481	3/31/2016 22:24	3/31/2016 22:45	3438	1220	152	Lincoln Ave & Diversey Pkwy	316	Damen Ave & Sunnyside Ave	Customer	NaN	NaN	22	4
77	9080463	3/31/2016 22:15	3/31/2016 22:17	3464	97	47	State St & Kinzie St	47	State St & Kinzie St	Customer	NaN	NaN	22	4
113	9080426	3/31/2016 22:02	3/31/2016 22:29	2716	1582	340	Clark St & Wrightwood Ave	161	Rush St & Superior St	Customer	NaN	NaN	22	4
115	9080424	3/31/2016 22:02	3/31/2016 22:29	1052	1611	340	Clark St & Wrightwood Ave	161	Rush St & Superior St	Customer	NaN	NaN	22	4

#对于多层分组后的数据，如果需要使用多重索引，此处需要传入一个元组
ridership_user_week.get_group(("Customer","3")).head()

	trip_id	starttime	stoptime	bikeid	tripduration	from_station_id	from_station_name	to_station_id	to_station_name	usertype	gender	birthyear	hour	weekday
7002	9071238	3/30/2016 23:57	3/31/2016 0:03	4018	317	305	Western Ave & Division St	374	Western Ave & Walton St	Customer	NaN	NaN	23	3
7007	9071233	3/30/2016 23:51	3/30/2016 23:56	4018	316	374	Western Ave & Walton St	305	Western Ave & Division St	Customer	NaN	NaN	23	3
7014	9071226	3/30/2016 23:28	3/30/2016 23:45	1142	993	71	Morgan St & Lake St	15	Racine Ave & 18th St	Customer	NaN	NaN	23	3
7117	9071031	3/30/2016 20:13	3/30/2016 20:34	1919	1302	284	Michigan Ave & Jackson Blvd	173	Mies van der Rohe Way & Chicago Ave	Customer	NaN	NaN	20	3
7118	9071030	3/30/2016 20:13	3/30/2016 20:34	2909	1312	284	Michigan Ave & Jackson Blvd	173	Mies van der Rohe Way & Chicago Ave	Customer	NaN	NaN	20	3

#对于分组后的数据，如果还需要进行一系列查询，可以使用apply方法，传入一个处理的函数
#假设我们需要查找上述分组后每组中用时最长的一组记录的起始和终止站点，可执行如下命令

max_time_from_id=ridership_user_week.apply(lambda record:record["from_station_id"][record["tripduration"].idxmax()])
max_time_to_id=ridership_user_week.apply(lambda record:record["to_station_id"][record["tripduration"].idxmax()])
max_time_time=ridership_user_week.apply(lambda record:record["tripduration"][record["tripduration"].idxmax()])
max_time_ridership=pd.DataFrame({"from_station_id":max_time_from_id,"to_station_id":max_time_to_id,
                                "tripduratio":max_time_time})

max_time_ridership

		from_station_id	to_station_id	tripduratio
usertype	weekday
Customer	3	35	277	81702
	4	340	229	68571
Subscriber	3	287	254	71758
	4	133	49	70977