Python for Finance: Analyze Big Financial Data

Out[57]:    array([‘2016-01-01’,    ‘2016-01-02’,   ‘2016-01-03’],  dtype=‘datetime64[D]’)

In  [ 58 ]: np.arange(‘2016-01-01’, ‘2016-10-01’,   dtype=‘datetime64[M]’)

#   monthly frequency

Out[58]:    array([‘2016-01’,   ‘2016-02’,  ‘2016-03’,  ‘2016-04’,  ‘2016-05’,  ‘2016-06’,

                                                                ‘2016-07’,  ‘2016-08’,  ‘2016-09’], dtype=‘datetime64[M]’)

In  [ 59 ]: np.arange(‘2016-01-01’, ‘2016-10-01’,   dtype=‘datetime64[W]’)[: 10 ]

#   weekly  frequency

Out[59]:    array([‘2015-12-31’,    ‘2016-01-07’,   ‘2016-01-14’,   ‘2016-01-21’,

                                                                ‘2016-01-28’,   ‘2016-02-04’,   ‘2016-02-11’,   ‘2016-02-18’,

                                                                ‘2016-02-25’,   ‘2016-03-03’],  dtype=‘datetime64[W]’)

You can also easily use subday frequencies, like hours or seconds (refer to the

documentation for all options):

In  [ 60 ]: dtl =   np.arange(‘2016-01-01T00:00:00’,    ‘2016-01-02T00:00:00’,

dtype=‘datetime64[h]’)

#   hourly  frequency

dtl[: 10 ]

Out[60]:    array([‘2016-01-01T00+0100’,    ‘2016-01-01T01+0100’,   ‘2016-01-01T02+0100’,

                                                                ‘2016-01-01T03+0100’,   ‘2016-01-01T04+0100’,   ‘2016-01-01T05+0100’,

                                                                ‘2016-01-01T06+0100’,   ‘2016-01-01T07+0100’,   ‘2016-01-01T08+0100’,

                                                                ‘2016-01-01T09+0100’],  dtype=‘datetime64[h]’)

Plotting date-time and/or time series data can sometimes be tricky. matplotlib has good

support for standard datetime objects. Transforming datetime64 information into

datetime information generally does the trick, as the following example, whose result is

shown in Figure C-1, illustrates:

In  [ 61 ]: import matplotlib.pyplot as plt

                                    %matplotlib inline

In  [ 62 ]: np.random.seed( 3000 )

rnd =   np.random.standard_normal(len(dtl)).cumsum()    **   2

In  [ 63 ]: fig =   plt.figure()

plt.plot(dtl.astype(dt.datetime),   rnd)

#   convert np.datetime to  datetime.datetime

plt.grid(True)

fig.autofmt_xdate()

#   autoformatting  of  datetime    x-ticks

Figure C-1. Plot with datetime.datetime x-ticks autoformatted

Finally, we also have an illustration of using arange with seconds and milliseconds as

frequencies:

In  [ 64 ]: np.arange(‘2016-01-01T00:00:00’,    ‘2016-01-02T00:00:00’,

dtype=‘datetime64[s]’)[: 10 ]

#   seconds as  frequency

Out[64]:    array([‘2016-01-01T00:00:00+0100’,  ‘2016-01-01T00:00:01+0100’,

                                                                ‘2016-01-01T00:00:02+0100’, ‘2016-01-01T00:00:03+0100’,

                                                                ‘2016-01-01T00:00:04+0100’, ‘2016-01-01T00:00:05+0100’,

                                                                ‘2016-01-01T00:00:06+0100’, ‘2016-01-01T00:00:07+0100’,