Cross-project Reopened Pull Request Prediction in GitHub

Abstract

This paper investigates cross-project prediction of reopened pull requests in GitHub repositories. It explores whether machine-learning models trained on data from one software project can predict reopened pull requests in another project, helping developers identify potentially problematic pull requests when sufficient project-specific historical data is unavailable.

Research Contributions

- Investigated the feasibility of predicting reopened pull requests across different GitHub projects.
- Evaluated the applicability of cross-project machine-learning models to pull request reopening prediction.
- Examined predictive features associated with reopened pull requests.
- Explored how cross-project prediction could support code review and pull request management when target-project training data is limited.

Methodology

The study applies empirical software engineering and supervised machine-learning techniques to pull request data collected from GitHub open-source projects. Cross-project prediction experiments are used to assess whether models trained on one project can generalize to another. The analysis examines predictive features and evaluates model performance in identifying pull requests likely to be reopened.

Keywords

How to Cite