Forecasting the number of end-of-life vehicles using a hybrid model based on grey model and artificial neural network.

Abstract This paper aims to better manage the reverse supply chain of the automotive industry in the context of green, circular, and sustainable development by predicting the number of end-of-life vehicles to be recycled through the establishment of a multi-factor model. The prediction of the number of end-of-life vehicles to be recycled in this paper will support the end-of-life vehicle recycling industry in terms of recycling management and investment decision-making and provide a reference for the formulation and implementation of policies relating to end-of-life vehicles. To solve the problems posed by nonlinear characteristics and uncertainty in the number of end-of-life vehicles recycled, and deal with the multiple factors influencing the recycling number, this paper presents a combined prediction model consisting of a grey model, exponential smoothing and an artificial neural network optimized by the particle swarm optimization (PSO) algorithm. Using Shanghai's end-of-life vehicle reverse logistics industry as an example, this study selects historical data about end-of-life vehicles recycled in Shanghai during the 2005–2016 period, identifies multiple influential factors, and validates the effectiveness and feasibility of the prediction model through empirical research. This paper proposes an effective prediction model for end-of-life vehicle industry managers, researchers, and regulators dealing with the industry's common challenges. Highlights • Hybrid prediction model based on grey model and artificial neural network predicts end-of-life vehicles recycling volume. • Hybrid prediction model improves prediction performance on data featuring randomness and small sample size. • Hybrid prediction model improves prediction performance of end-of-life vehicles recycling volume affected by multi-factors. • Artificial neural network prediction model using particle swarm optimization algorithm has better prediction performance. • Model proposed in this paper is tested by application to real-life data from the reverse logistics industry in Shanghai. [ABSTRACT FROM AUTHOR]