A Data-Driven Software Configuration Management System Based on Knowledge Graph and Graph Neural Network

As embedded and mechatronic systems grow increasingly complex, frequent software version updates and the ever-expanding number of configuration items have exposed traditional software configuration management to challenges such as delayed information synchronization and high manual maintenance costs. This paper proposes a big data-based software configuration management system that leverages knowledge graphs and Graph Neural Network (GNN) technology to model configuration item relationships, analyze change impacts, and infer verification requirements. The system constructs a heterogeneous graph structure comprising three node types: "configuration items,""change records,"and "verification requirements."By learning dependencies between different configuration items through GNNs, it enables automatic synchronization of software entry/exit information and intelligent push notifications for change recommendations. Integrated as a plugin into existing configuration management platforms, the system delivers change identification, information synchronization, and validation recommendation capabilities. Case studies across multi-model embedded software management scenarios demonstrate its effectiveness in enhancing change processing efficiency and configuration consistency, providing technical support for comprehensive software lifecycle management.