DL-SPP: A DEEP LEARNING SYSTEM FOR SPARE PARTS PREDICTION IN FIELD SERVICE OPERATIONS

Accurate spare parts prediction is crucial in field service operations for the maintenance and repair of household appliances, where missing components frequently lead to repeated technician visits, increased costs, and diminished service quality. In this context, we present the DL-SPP (Deep Learning Spare Part Prediction) system that predicts the set of spare parts likely needed before the technician’s intervention, using both structured appliance metadata and semantic representations of fault descriptions. The prediction task is framed as a multi-label classification problem, where the model learns to associate appliance types and fault reports with the components used in past repairs through a transformer-based architecture. Evaluated on a real-world dataset of repair tasks, DL-SPP achieves up to a 56% reduction in repeated field service technician visits, outperforming a state-of-the-art baseline that achieves 50%. These results highlight the DL-SPP system as a scalable and effective solution for enhancing repair planning and inventory optimization in operational field service contexts.