Share
Export Citation
KG-GNNRec: A Knowledge Graph and Graph Neural Network-Based Framework for Automotive Spare Parts Maintenance Recommendation
Usman S.
IEEE International Conference on Communication Networks and Satellite Comnetsat
Abstract
The increasing complexity of modern vehicles and the demand for efficient after-sales services highlight the importance of predictive maintenance strategies that can anticipate potential failures before they occur. Conventional maintenance approaches in automotive workshops are often reactive, conducted only after breakdowns, resulting in higher costs and longer downtimes. To overcome these limitations, this study proposes a Knowledge Graph–based Graph Neural Network Recommendation framework (KG-GNNRec) that leverages vehicle service history data and interrelated knowledge among components to provide proactive spare-part maintenance recommendations. The proposed framework constructs a Knowledge Graph (KG) to represent relationships among entities such as vehicles, subsystems, components, and maintenance events. A Graph Neural Network (GNN) module is then employed to perform attention-based embedding propagation, enabling the model to capture high order dependencies and semantic relations within complex automotive maintenance data. This integration allows the system to generate more accurate and explainable recommendations compared to traditional, feature-based approaches. Experimental evaluation using real workshop service records demonstrates that KG-GNNRec effectively improves recommendation accuracy and interpretability. The model can identify relevant component relationships and maintenance patterns, supporting data-driven decision-making in vehicle servicing. The results indicate that the proposed approach contributes to reducing unplanned downtime, improving maintenance efficiency, and enhancing customer satisfaction through intelligent and interpretable recommendations.