Spare part demand forecasting using PSO trained Quantile Regression Neural Network

The after-sales service is a crucial component within the supply chain. Rapid upgrade of electronic product parts leads to the inability of part suppliers to maintain production lines all the time. After-sale service centers need to forecast the volume of the required repair parts to satisfy the needs of customers. In this paper, the demand for spare parts is forecast by considering both regular monthly and Last Time Buy replenishment volumes until the end of the product warranty period. Given the proven effectiveness of Quantile Regression Neural Network (QRNN) and Recurrent Neural Network (RNN) in time-series forecasting, this paper suggests a hybrid network structure combining QRNN and RNN for forecasting spare part demand. Furthermore, an improved Particle Swarm Optimization (PSO) method is designed to optimize the network training process. Real-world cases involving different categories of spare parts consumption, where the results demonstrate the effectiveness of the tailored mechanisms, such as RNN structure and PSO-inspired network training. Moreover, our proposed algorithm demonstrates better performance compared to the state-of-the-art algorithms in terms of six standard point forecast error metrics.