Expanding window operations in Pandas

Introduction

Expanding window operations help you calculate running totals or averages that grow as you move through your data. This shows how values change over time with all past data included.

Tracking cumulative sales or revenue over days or months.

Calculating a running average of temperatures to see trends.

Monitoring total website visits up to each day.

Analyzing cumulative scores in a game or competition.

Measuring total rainfall accumulated over a season.

Syntax

Pandas

DataFrame.expanding(min_periods=1).function()

min_periods sets the minimum number of data points needed before calculating the result.

Common functions include sum(), mean(), max(), and min().

Examples

Calculate the cumulative sum of the 'value' column, starting from the first row.

Pandas

df['value'].expanding().sum()

Calculate the expanding mean but only start showing results after at least 3 data points.

Pandas

df['value'].expanding(min_periods=3).mean()

Find the maximum value so far for each column in the DataFrame.

Pandas

df.expanding().max()

Sample Program

This code creates a small sales dataset over 5 days. It then calculates the total sales up to each day and the average sales starting from day 2.

Pandas

import pandas as pd

data = {'day': [1, 2, 3, 4, 5], 'sales': [100, 150, 200, 130, 170]}
df = pd.DataFrame(data)

# Calculate cumulative sales using expanding sum
df['cumulative_sales'] = df['sales'].expanding().sum()

# Calculate expanding average sales with minimum 2 days
df['expanding_avg'] = df['sales'].expanding(min_periods=2).mean()

print(df)

Important Notes

If min_periods is not set, it defaults to 1, so calculations start immediately.

Expanding windows always start from the first row and include all previous rows.

Use expanding operations to see how totals or averages grow over time.

Summary

Expanding window operations calculate cumulative values including all past data.

They are useful for running totals, averages, and tracking changes over time.

Set min_periods to control when results start appearing.