学习python需要熟记一些语法，然后去灵活应用。有些知识虽然前期没学习，但有时模块可以分开去学习掌握。我按视频学习的pandas 时间序列，有不太清楚地方，目前只是把听后差不多理解的笔记整理了， 不清楚处未做整理。

pandas 时间序列

#导包

import numpy as np

import pandas as pd
pd.Timestamp('2030-2-3') #时刻数据

pd.Period('2030-2-3',freq='D') #时期数据
#批量生成时刻数据
index=pd.date_range('2030.02.13',periods=4,freq='D')
# index=pd.period_range('2030.02.13',periods=4,freq='D')
index
#freq:Y:年，M：月，D：日

DatetimeIndex(['2030-02-13', '2030-02-14', '2030-02-15', '2030-02-16'], dtype='datetime64[ns]', freq='D')

#时间戳索引
pd.Series(np.random.randint(0,10,size=4),index=index)

2030-02-13 7

2030-02-14 3

2030-02-15 2

2030-02-16 6

Freq: D, dtype: int32

#转换方法
pd.to_datetime(['2030-3-14','2030.03.14','14/03/2030','2030/3/14'])

DatetimeIndex(['2030-03-14', '2030-03-14', '2030-03-14', '2030-03-14'], dtype='datetime64[ns]', freq=None)

dt=pd.to_datetime([20300314],unit='s')
dt

DatetimeIndex(['1970-08-23 22:58:34'], dtype='datetime64[ns]', freq=None)

#时间差 DateOffset
dt+pd.DateOffset(hours=8)

DatetimeIndex(['1970-08-24 06:58:34'], dtype='datetime64[ns]', freq=None)

dt+pd.DateOffset(days=-8)

DatetimeIndex(['1970-08-15 22:58:34'], dtype='datetime64[ns]', freq=None)

#时间戳的索引和切片
index=pd.bdate_range('2030-3-14',periods=100,freq='D')
index

DatetimeIndex(['2030-03-14', '2030-03-15', '2030-03-16', '2030-03-17',

'2030-03-18', '2030-03-19', '2030-03-20', '2030-03-21',

'2030-03-22', '2030-03-23', '2030-03-24', '2030-03-25',

'2030-03-26', '2030-03-27', '2030-03-28', '2030-03-29',

'2030-03-30', '2030-03-31', '2030-04-01', '2030-04-02',

'2030-04-03', '2030-04-04', '2030-04-05', '2030-04-06',

'2030-04-07', '2030-04-08', '2030-04-09', '2030-04-10',

'2030-04-11', '2030-04-12', '2030-04-13', '2030-04-14',

'2030-04-15', '2030-04-16', '2030-04-17', '2030-04-18',

'2030-04-19', '2030-04-20', '2030-04-21', '2030-04-22',

'2030-04-23', '2030-04-24', '2030-04-25', '2030-04-26',

'2030-04-27', '2030-04-28', '2030-04-29', '2030-04-30',

'2030-05-01', '2030-05-02', '2030-05-03', '2030-05-04',

'2030-05-05', '2030-05-06', '2030-05-07', '2030-05-08',

'2030-05-09', '2030-05-10', '2030-05-11', '2030-05-12',

'2030-05-13', '2030-05-14', '2030-05-15', '2030-05-16',

'2030-05-17', '2030-05-18', '2030-05-19', '2030-05-20',

'2030-05-21', '2030-05-22', '2030-05-23', '2030-05-24',

'2030-05-25', '2030-05-26', '2030-05-27', '2030-05-28',

'2030-05-29', '2030-05-30', '2030-05-31', '2030-06-01',

'2030-06-02', '2030-06-03', '2030-06-04', '2030-06-05',

'2030-06-06', '2030-06-07', '2030-06-08', '2030-06-09',

'2030-06-10', '2030-06-11', '2030-06-12', '2030-06-13',

'2030-06-14', '2030-06-15', '2030-06-16', '2030-06-17',

'2030-06-18', '2030-06-19', '2030-06-20', '2030-06-21'],

dtype='datetime64[ns]', freq='D')

ts=pd.Series(range(len(index)),index=index)
ts

2030-03-14 0

2030-03-15 1

2030-03-16 2

2030-03-17 3

2030-03-18 4

..

2030-06-17 95

2030-06-18 96

2030-06-19 97

2030-06-20 98

2030-06-21 99

Freq: D, Length: 100, dtype: int64

#切片
ts['2030-03-15':'2030-03-22']

2030-03-15 1

2030-03-16 2

2030-03-17 3

2030-03-18 4

2030-03-19 5

2030-03-20 6

2030-03-21 7

2030-03-22 8

Freq: D, dtype: int64

ts[pd.date_range('2030-3-24',periods=10,freq='D')]

2030-03-24 10

2030-03-25 11

2030-03-26 12

2030-03-27 13

2030-03-28 14

2030-03-29 15

2030-03-30 16

2030-03-31 17

2030-04-01 18

2030-04-02 19

Freq: D, dtype: int64

#属性
ts.index
ts.index.year   #年
ts.index.month  #月
ts.index.day  #日
ts.index.dayofweek   #星期几

Int64Index([3, 4, 5, 6, 0, 1, 2, 3, 4, 5, 6, 0, 1, 2, 3, 4, 5, 6, 0, 1, 2, 3,

4, 5, 6, 0, 1, 2, 3, 4, 5, 6, 0, 1, 2, 3, 4, 5, 6, 0, 1, 2, 3, 4,

5, 6, 0, 1, 2, 3, 4, 5, 6, 0, 1, 2, 3, 4, 5, 6, 0, 1, 2, 3, 4, 5,

6, 0, 1, 2, 3, 4, 5, 6, 0, 1, 2, 3, 4, 5, 6, 0, 1, 2, 3, 4, 5, 6,

0, 1, 2, 3, 4, 5, 6, 0, 1, 2, 3, 4],

dtype='int64')

度进行数据聚合

ts.resample('2D').sunm() #以2天为单位进行汇总 求和

ts.resample('2W').sunm() #以2周为单位进行汇总 求和

ts.resample('3M').sunm() #以3个月为单位进行汇总 求和

ts.resample('H').sunm() #以1小时为单位进行汇总 求和

ts.resample('T').sunm() #以1分钟为单位进行汇总 求和

#DataFrame重采样

d={'price':[10,11,2,44,55,66],'score':[40,30,20,50,60,70,80,10],'week':[pd.data_range('2030-3-1',periods=8,freq='w')}

df=pd.DataFrame(d)

df