KG-Commit: Knowledge-Graph Dumps for Just-in-Time Defect Prediction on Eleven Apache Projects

Neo4j knowledge-graph dumps accompanying the paper "KG-Commit: A DynamicKnowledge Graph for Online Just-in-Time Software Defect Prediction". Contains one Neo4j dump per Apache project (ActiveMQ, Camel, Cassandra, Flink,Groovy, HBase, Hive, Kafka, Spark, Zeppelin, ZooKeeper), each holding the fullybuilt three-layer knowledge graph: Core (repository entities), AST (within-filestructure with deltas), and CSTG (commit-message and code-change semantics).Loading a dump reproduces the graph state used for inference in the paper,without rebuilding. Requires Neo4j 2026.07.0 or newer (Java 21). These dumps carry store kernelversion prefix 29 and cannot be loaded by Neo4j 5.x, including 5.26.x. See README.md for the file-to-project mapping, sizes, and loading instructions. Code: https://github.com/Knowledge4Software/KG-CommitCommit labels derive from ApacheJIT (Keshavarz & Nagappan, 2022):https://doi.org/10.5281/zenodo.5907001