Process metrics over time: replication package for a snapshot study of four open-source projects

Code, derived data and results for the article "Predicting Defect-Prone Files from Process Metrics over Time: A Snapshot Study of Four Open-Source Projects". At half-year snapshots from 2019 to 2025, process metrics are computed for each production code file of Kafka, Hadoop, Spring Boot and Elasticsearch (Express is described but not modelled), using only commits before the snapshot. The label records whether a bug-fixing commit touches the file in the next 180 days. Models are trained on earlier snapshots and tested on the next one. The package contains the mining, labelling, snapshot, evaluation and statistics scripts, the derived data, all result tables and figures, a manually coded validation sample of 195 commits, and a sensitivity check with keyword-only labels. Every number in the article is produced by these scripts. See README.md for the pipeline and how to rerun it. Code: MIT. Data and results: CC BY 4.0.