An Exploratory Study of Project Activity Changepoints in Open Source Software Evolution

TL;DR

This study analyzes open source project activity to identify the frequency and nature of abrupt changes, revealing that most projects experience several small to moderate activity shifts with some larger fluctuations.

Contribution

It applies the PELT changepoint detection algorithm to a large dataset of open source projects, providing new insights into activity change patterns.

Findings

Over 90% of projects have 1-6 changepoints.

Activity increases and decreases occur with similar frequency.

Most changes are small, but some are significantly large.

Abstract

To explore the prevalence of abrupt changes (changepoints) in open source project activity, we assembled a dataset of 8,919 projects from the World of Code. Projects were selected based on age, number of commits, and number of authors. Using the nonparametric PELT algorithm, we identified changepoints in project activity time series, finding that more than 90% of projects had between one and six changepoints. Increases and decreases in project activity occurred with roughly equal frequency. While most changes are relatively small, on the order of a few authors or few dozen commits per month, there were long tails of much larger project activity changes. In future work, we plan to focus on larger changes to search for common open source lifecycle patterns as well as common responses to external events.