Working with Time Series

Creating Time Series

Here is how to create a simple Time Series.

import pandas as pd
import numpy as np

# The number of sample to generate
nb_sample = 100

# Seeding to obtain a reproductible dataset
np.random.seed(0)

se = pd.Series(np.random.randint(0, 100, nb_sample),
                  index = pd.date_range(start = pd.to_datetime('2016-09-24'),
                                        periods = nb_sample, freq='D'))
se.head(2)

# 2016-09-24    44
# 2016-09-25    47

se.tail(2)

# 2016-12-31    85
# 2017-01-01    48

Partial String Indexing

A very handy way to subset Time Series is to use partial string indexing. It permits to select range of dates with a clear syntax.

Getting Data

We are using the dataset in the Creating Time Series example

Displaying head and tail to see the boundaries

se.head(2).append(se.tail(2))

# 2016-09-24    44
# 2016-09-25    47
# 2016-12-31    85
# 2017-01-01    48

Subsetting

Now we can subset by year, month, day very intuitively.

By year

se['2017']

# 2017-01-01    48

By month

se['2017-01']

# 2017-01-01    48

By day

se['2017-01-01']

# 48

With a range of year, month, day according to your needs.

se['2016-12-31':'2017-01-01']

# 2016-12-31    85
# 2017-01-01    48

pandas also provides a dedicated truncate function for this usage through the after and before parameters — but I think it’s less clear.

se.truncate(before='2017')

# 2017-01-01    48

se.truncate(before='2016-12-30', after='2016-12-31')

# 2016-12-30    13
# 2016-12-31    85