EBC software evolution: analysis dataset and replication code for 337 Java projects

This package contains the complete analysis-ready dataset for 337 open-source Java projects through 31 August 2026: 717,648 selected first-parent commits, 454,524 distinct project-file identities, saved EBC predictions and 3,584,264 file-change events. It includes mining, classification and metric-processing source code, analytical exports, reported figures, and portable reproduction scripts. The main temporal analysis uses commit progression. Alternative timing views and exploratory analyses are separately labelled. The package enables reanalysis from saved predictions. Original Java source/Git archives and the trained classification model are not redistributed; fresh classification needs these separately. Creator identity is withheld for review. Data are CC0; author-owned code is MIT; third-party components retain their own terms.