[Paper Review] On the Popularity of GitHub Applications: A Preliminary Note

Hudson Silva Borges, Marco Túlio Valente+2 more|arXiv (Cornell University)|Jul 2, 2015

Advanced Malware Detection Techniques6 references18 citations

TL;DR

This paper proposes a framework to assess GitHub software popularity using star counts and identifies four growth patterns—sustainable, fast, slow, and viral—based on longitudinal star trends. It finds strong correlations between stars, forks, and client usage (e.g., Spearman’s rho = 0.68 with NPM dependents), validating stars as a reliable popularity metric, with JavaScript dominating the most popular repositories.

ABSTRACT

GitHub is the world's largest collection of open source software. Therefore, it is important both to software developers and users to compare and track the popularity of GitHub repositories. In this paper, we propose a framework to assess the popularity of GitHub software, using their number of stars. We also propose a set of popularity growth patterns, which describe the evolution of the number of stars of a system over time. We show that stars tend to correlate with other measures, like forks, and with the effective usage of GitHub software by third-party programs. Throughout the paper we illustrate the application of our framework using real data extracted from GitHub.

Motivation & Objective

• To develop a framework for assessing the popularity of open-source software hosted on GitHub.
• To define and classify distinct patterns of popularity growth over time using star count evolution.
• To validate the star count as a reliable proxy for real-world popularity by correlating it with forks and client usage.
• To analyze the distribution of popular and very popular repositories across programming languages.
• To provide actionable insights for developers and users on tracking software adoption and maturity.

Proposed method

• Define 'popular' as the top 10% of repositories by star count (≥1,459 stars on May 1, 2015) and 'very popular' as the top 1% (≥19,570 stars).
• Use logarithmic rank transformation (base 2) to normalize star-based rankings and reduce skew in high-impact repositories.
• Classify growth patterns based on rank variation: sustainable (|R_BOTTOM - R_TOP| < 0.25), fast, slow, and viral, using 52-week historical data.
• Correlate star counts with forks and client usage (via NPM API for Node.js modules) to validate popularity metrics.
• Analyze a snapshot of 24,000 repositories across 24 top programming languages from GitHub’s May 1, 2015 data.
• Apply statistical methods including Spearman’s rank correlation to assess relationships between popularity indicators.

Experimental results

Research questions

• RQ1 How can we systematically classify the popularity of GitHub repositories using star count trends over time?
• RQ2 What are the dominant patterns of popularity growth (e.g., sustainable, viral) in open-source software on GitHub?
• RQ3 To what extent do star counts correlate with other indicators like forks and client usage?
• RQ4 Which programming languages dominate the most popular and very popular repositories on GitHub?
• RQ5 Is the number of stars a reliable proxy for real-world usage and adoption of GitHub-hosted software?

Key findings

• JavaScript dominates the most popular GitHub repositories, with 35% of very popular systems and 46% of all popular systems.
• 22% of popular systems (466 out of 2,138) exhibit sustainable growth, maintaining stable rankings over one year.
• 5% of popular systems show fast growth, while less than 1% show slow growth, and 37 systems exhibit viral growth patterns.
• There is a strong positive correlation (Spearman’s rho = 0.68) between the number of stars and the number of dependents (clients) for Node.js libraries on NPM.
• The number of stars correlates significantly with the number of forks, reinforcing its validity as a popularity metric.
• The top three repositories—twbs/bootstrap, angular/angular.js, and mbostock/d3—accounted for the highest star counts, with over 100,000 stars each.