第十章 时间序列

from datetime import datetime

now=datetime.now()

now
Out[3]: datetime.datetime(2017, 6, 28, 9, 15, 30, 916000)

now.year,now.month,now.day
Out[4]: (2017, 6, 28)

delta=datetime(2017,6,28)-datetime(2008,6,24,8,15)

delta
Out[6]: datetime.timedelta(3290, 56700)

delta.days
Out[8]: 3290
delta.seconds
Out[9]: 56700
from datetime import timedelta

delta=datetime(2017,9,2)-datetime(2017,6,28,9,26)

delta
Out[12]: datetime.timedelta(65, 52440)
delta=datetime(2017,9,2)-datetime(2017,6,28,19,26)

delta
Out[14]: datetime.timedelta(65, 16440)
#给datetime对象加上（或减去）一个或多个timedelta
from datetime import timedelta
start=datetime(2017,6,28)

start+timedelta(65)
Out[18]: datetime.datetime(2017, 9, 1, 0, 0)
#字符串与datetime的相互转换
stamp=datetime(2017,6,28)

str(stamp)
Out[20]: '2017-06-28 00:00:00'
stamp.strftime('%Y-%m-%d')
Out[21]: '2017-06-28'

value='2017-06-28'

datetime.strptime(value,'%Y-%m-%d')
Out[24]: datetime.datetime(2017, 6, 28, 0, 0)

datestrs=['7/6/2011','8/6/2011']

[datetime.strptime(x,'%m/%d/%Y') for x in datestrs]
Out[29]: [datetime.datetime(2011, 7, 6, 0, 0), datetime.datetime(2011, 8, 6, 0, 0)]
from dateutil.parser import parse

parse('2017-06-28')
Out[31]: datetime.datetime(2017, 6, 28, 0, 0)

parse('Jun 28,2017 9:45 PM')
Out[32]: datetime.datetime(2017, 6, 28, 21, 45)
#国际通用格式：日在月前面。传入dayfirst=True即可
parse('6/12/2011',dayfirst=True)
Out[33]: datetime.datetime(2011, 12, 6, 0, 0)
datestrs
Out[34]: ['7/6/2011', '8/6/2011']
import pandas as pd

pd.to_datetime(datestrs)
Out[36]: DatetimeIndex(['2011-07-06', '2011-08-06'], dtype='datetime64[ns]', freq=None)
idx=pd.to_datetime(datestrs+[None])

idx
Out[38]: DatetimeIndex(['2011-07-06', '2011-08-06', 'NaT'], dtype='datetime64[ns]', freq=None)

idx[2]
Out[39]: NaT
#NaT是pandas中时间戳的缺失值
pd.isnull(idx)
Out[40]: array([False, False,  True], dtype=bool)
#时间序列基础
from datetime import datetime
import numpy as np
from pandas import Series
dates=[datetime(2011,1,2),datetime(2011,1,5),datetime(2011,1,7),
       datetime(2011,1,8),datetime(2011,1,10),datetime(2011,1,12)]
ts=Series(np.random.randn(6),index=dates)
ts
Out[47]: 
2011-01-02    1.582904
2011-01-05   -1.439438
2011-01-07   -0.411849
2011-01-08    0.747215
2011-01-10    0.384336
2011-01-12    0.432144
dtype: float64
type(ts)
Out[48]: pandas.core.series.Series
ts.index
Out[49]: 
DatetimeIndex(['2011-01-02', '2011-01-05', '2011-01-07', '2011-01-08',
               '2011-01-10', '2011-01-12'],
              dtype='datetime64[ns]', freq=None)
ts+ts[::2]
Out[50]: 
2011-01-02    3.165807
2011-01-05         NaN
2011-01-07   -0.823697
2011-01-08         NaN
2011-01-10    0.768672
2011-01-12         NaN
dtype: float64
#DatetimeIndex的各个标量值是timastamp的对象
stamp=ts.index[0]
stamp
Out[53]: Timestamp('2011-01-02 00:00:00')
#索引、选取、子集的构造
#TimeSeries是Series的一个子类，在索引以及数据选取方面用法相同
stamp=ts.index[2]
ts[stamp]
Out[60]: -0.41184861375211623
#传入一个可被解释为日期的字符串
ts['1/10/2011']
Out[62]: 0.38433616480100646
#对于较长的时间序列，只需传入“年”或“年月”即可轻松选取数据的切片
longer_ts=Series(np.random.randn(1000),index=pd.date_range('1/1/2000',periods=1000))
longer_ts
Out[65]: 
2000-01-01   -0.358478
2000-01-02   -0.004255
2000-01-03   -0.051692
2000-01-04    0.704217
2000-01-05    0.489014
2000-01-06   -0.905860
2000-01-07    0.523610
2000-01-08    0.075787
2000-01-09    0.530821
2000-01-10    0.909307
2000-01-11   -0.173890
2000-01-12    0.159520
2000-01-13   -1.144813
2000-01-14    0.549058
2000-01-15   -0.474099
2000-01-16   -1.272199
2000-01-17    0.818596
2000-01-18   -1.209547
2000-01-19    1.403011
2000-01-20    1.430430
2000-01-21    1.203109
2000-01-22    0.830908
2000-01-23    0.100247
2000-01-24    0.466166
2000-01-25    0.548837
2000-01-26    0.328641
2000-01-27    0.485374
2000-01-28   -0.829345
2000-01-29   -0.730873
2000-01-30   -0.196522
  
2002-08-28    1.224319
2002-08-29   -0.250930
2002-08-30   -0.543701
2002-08-31   -0.072269
2002-09-01   -0.122738
2002-09-02    0.917622
2002-09-03    0.814192
2002-09-04   -0.130414
2002-09-05   -0.368046
2002-09-06    0.552689
2002-09-07    2.261430
2002-09-08   -1.050885
2002-09-09   -0.015522
2002-09-10    0.560897
2002-09-11   -0.069163
2002-09-12   -0.432103
2002-09-13   -0.847568
2002-09-14    1.127829
2002-09-15    0.795124
2002-09-16   -1.620212
2002-09-17    0.307566
2002-09-18    0.159686
2002-09-19    0.491870
2002-09-20   -0.760113
2002-09-21   -1.108078
2002-09-22   -1.078166
2002-09-23    1.054080
2002-09-24    1.758295
2002-09-25    0.283322
2002-09-26    0.416153
Freq: D, dtype: float64
#通过日期进行切片的方式只对规则Series有效
ts[datetime(2011,1,7):]
2011-01-07   -0.411849
2011-01-08    0.747215
2011-01-10    0.384336
2011-01-12    0.432144
dtype: float64
#也可用不存在该时间序列中的时间戳对其进行切片

ts['1/6/2011':'1/11/2011']
Out[69]: 
2011-01-07   -0.411849
2011-01-08    0.747215
2011-01-10    0.384336
dtype: float64
ts.truncate(after='1/9/2011')
Out[70]: 
2011-01-02    1.582904
2011-01-05   -1.439438
2011-01-07   -0.411849
2011-01-08    0.747215
dtype: float64
from pandas import DataFrame
dates=pd.date_range('1/1/2000',periods=100,freq='W-WED')
long_df=DataFrame(np.random.randn(100,4),index=dates,columns=['Colorado','Texas','New York','Ohio'])
long_df.ix['6-2001']
            Colorado     Texas  New York      Ohio
2001-06-06  0.663364 -0.225784 -0.397083  0.270297
2001-06-13  0.978811  1.239237  1.860109  0.322908
2001-06-20 -1.187403  1.011756  0.817634  0.503117
2001-06-27 -0.110130  0.367503  1.627412  0.758577