Research on machine learning algorithm for intelligent prediction of automotive parts demand

This paper aims to design and implement an efficient auto parts demand forecasting system, with the manufacturing plant as the core, and conducts in-depth research on the influencing factors of demand in the auto parts supply chain and main business processes. Traditional demand forecasting often relies only on local data and fails to consider the mutual influence between nodes in the entire supply chain, resulting in inaccurate predictions. Thus, this paper conducts a comparative analysis of auto parts forecasting methods from both domestic and international perspectives. It selects the BP neural network model, the BP neural network model optimized by the bat algorithm (BA), and the GA-BP artificial neural network model. The effectiveness of these models is verified through training and testing. The system development utilizes the JAWS framework to facilitate secure data exchange between enterprises and combines the RSA encryption algorithm to ensure data integrity. While the BP artificial neural network model performs well, it tends to get trapped in local optima and lacks stability. In contrast, the GA-BP artificial neural network model and the BA-BP neural network model demonstrate significantly superior performance, featuring smaller prediction errors and a stronger ability to analyze data trends and internal relationships, thereby showing notable advantages in auto parts demand forecasting. By effectively integrating data and optimizing algorithms, this paper provides a new approach and tool for auto parts demand forecasting, and proposes suggestions for future research directions, including further exploring more efficient data preprocessing methods and model optimization strategies to better serve the intelligent development of the auto parts industry.