Module 5 - Tutorial - Time Series

import pandas as pd

import numpy as np

import seaborn; seaborn.set()

import matplotlib.pyplot as plt

from datetime import datetime

%matplotlib inline

date = np.array('2015-07-04', dtype=np.datetime64)

date

#creates 12 concsecutive dates starting from July 4, 2015

date + np.arange(12)

#day-based datetime

np.datetime64('2015-07-04')

#minute-based datetime

#sets to time zone of local computer

np.datetime64('2015-07-04 12:00')

date = pd.to_datetime("4th of July, 2015")

date

#output day of the week

date.strftime('%A')

#create 12 consecutive dates starting from July 4, 2015

date + pd.to_timedelta(np.arange(12), 'D')

#make index using timestamp

index = pd.DatetimeIndex(['2014-07-04', '2014-08-04',

                          '2015-07-04', '2015-08-04'])

data = pd.Series([0,1,2,3], index=index) #a series is a singular column from a dataframe

data

#can slice date index similarly to regular index slicing

#includes ending index

data['2014-07-04':'2015-07-04']

#get rows from 2015

data['2015']

#will convert different formats into datetime

dates = pd.to_datetime([datetime(2015, 7, 3), '4th of July 2015', '2015-Jul-6', '07-07-2015', '20150708'])

dates

#assigned a 'day' frequency

dates.to_period('D')

#calculate # of days between 07-03-2015 and each date

dates - dates[0]

#create days between start date and end date

#default frequency is 'day'

pd.date_range('2015-07-03', '2015-07-10')

#create days from start date with periods (how many)

pd.date_range('2015-07-03', periods=8)

#make range using hourly frequency

#default start at 00:00 (midnight)

pd.date_range('2015-07-03', periods=8, freq='H')

#create period frequency by month

pd.period_range('2015-07', periods=8, freq='M')

#create only hour frequencies (no date attached)

#seconds also included

pd.timedelta_range(0, periods=10, freq='H')

#create 2 hour 30 minute intervals

pd.timedelta_range(0, periods=9, freq='2H30T')

#create business day offsets

from pandas.tseries.offsets import BDay

pd.date_range('2015-07-01', periods=5, freq=BDay())

#make dates the index

#format dates as we load the data

filename = FremontBridgeBicycle.csv

data = pd.read_csv(filename, index_col="Date", parse_dates=True)

data.head()

data.describe()

#rename columns with shorter name

#create a new column 'Total' with combined values of west and east columns

data.columns = ['West', 'East']

data['Total'] = data.eval('West + East')

#bar plot of data

data.plot()

plt.ylabel('Hourly Bicycle Count')

#hourly intervals are too narrow to make sense of data

#aggregate as weekly data instead

weekly = data.resample('W').sum()

weekly.plot(style=[':', '--', '-']) #line styles for each feature

plt.ylabel('Weekly Bicycle Count')

#see what an average day looks like

by_time = data.groupby(data.index.time).mean()

hourly_ticks = 4*60*60*np.arange(6) #6 intervals of 4 hours

by_time.plot(xticks=hourly_ticks, style=[':','--','-'])

#What does bike usage look like by day of the week?

by_weekday = data.groupby(data.index.dayofweek).mean()

by_weekday.index = ['Mon', 'Tue', 'Wed', 'Thur', 'Fri', 'Sat', 'Sun']

by_weekday.plot(style=[':','--','-'])

#set conditions to show hourly trend on weekdays vs weekends

weekend = np.where(data.index.weekday < 5, 'Weekday', 'Weekend')

by_time = data.groupby([weekend, data.index.time]).mean()

fig, ax = plt.subplots(1, 2, figsize=(14, 5))

by_time.loc['Weekday'].plot(ax=ax[0], title='Weekdays',

                           xticks=hourly_ticks, style=[':', '--', '-'])

by_time.loc['Weekend'].plot(ax=ax[1], title='Weekends',

                           xticks=hourly_ticks, style=[':', '--', '-']);