Window Operations—Rolling

Categories of windowing operations

Windowing operations are one of the features in pandas that stand at the intersection of grouping and aggregation. The idea is to divide the data into partitions (aka windows of time) and then calculate an aggregate function for each window instead of the entire dataset.

There are four categories of windowing operations, namely:

• Rolling windows

• Weighted windows

• Expanding windows

• Exponential moving windows

We’ll use the Amazon (AMZN) stock prices in 2017 as the primary dataset for the code examples in this lesson.

Note: Windowing operations currently only support numeric data (integer and float) and will always return float64 values.

Rolling windows

We start by looking at rolling windows that are characterized by a fixed window size that slides along the data points. Each window contains the current observation at a time point and a fixed number of preceding observations. Rolling windows are useful for smoothing out short-term fluctuations and highlighting longer-term trends in the data.

The image below provides a simple illustration of what rolling windows of a fixed size three look like:

The rolling() method

We can use the rolling() method for the computation of rolling statistics on our time series data. For example, we can use the following code to calculate the rolling mean of the closing price over a ...