Window functions (expanding, ewm) in Data Analysis Python

Introduction

Window functions help us look at data in small parts or over time. They let us calculate things like averages that change as we move through the data.

To see how the average of sales changes day by day in a store.

To smooth noisy data like stock prices to see trends better.

To calculate running totals or averages in a list of numbers.

To analyze time series data where recent points matter more.

To compare each data point with all previous points in order.

Syntax

Data Analysis Python

df['column'].expanding().mean()
df['column'].ewm(span=3).mean()

expanding() looks at all data from the start up to the current point.

ewm() means Exponentially Weighted Moving, which gives more weight to recent data.

Examples

This calculates the average from the first value up to each point.

Data Analysis Python

import pandas as pd

data = [1, 2, 3, 4, 5]
df = pd.DataFrame({'values': data})

# Expanding mean
exp_mean = df['values'].expanding().mean()
print(exp_mean)

This calculates a weighted average giving more importance to recent values.

Data Analysis Python

import pandas as pd

data = [1, 2, 3, 4, 5]
df = pd.DataFrame({'values': data})

# Exponentially weighted mean with span=2
ewm_mean = df['values'].ewm(span=2).mean()
print(ewm_mean)

Sample Program

This program shows how to calculate the expanding and exponentially weighted averages of sales over dates.

Data Analysis Python

import pandas as pd

# Create sample data
sales = [10, 20, 15, 30, 25]
dates = pd.date_range('2024-01-01', periods=5)
df = pd.DataFrame({'date': dates, 'sales': sales})
df.set_index('date', inplace=True)

# Calculate expanding mean of sales
expanding_avg = df['sales'].expanding().mean()

# Calculate exponentially weighted mean with span=3
ewm_avg = df['sales'].ewm(span=3).mean()

print('Expanding mean:')
print(expanding_avg)
print('\nExponentially weighted mean:')
print(ewm_avg)

Important Notes

Expanding windows always start from the first data point and grow with each step.

Exponentially weighted means react faster to recent changes than simple averages.

You can adjust the span in ewm() to control how much recent data matters.

Summary

Window functions help analyze data step-by-step or over time.

expanding() looks at all data up to now.

ewm() gives more weight to recent data for smoother trends.