Research on E-Commerce Product Demand Forecasting Method Based on ARIMA-SVR-PSO

With the rapid development of e-commerce, accurate forecasting of product demand is crucial for optimizing inventory management and enhancing customer satisfaction. This paper presents an e-commerce product demand forecasting method based on ARIMA-SVR-PSO. The method begins with data preprocessing and statistical analysis to construct datasets that include historical shipment volumes, product information, merchant information, and warehouse information. It then utilizes intermittent demand time series forecasting techniques, such as exponential smoothing and the Croston method, for preliminary demand predictions. Furthermore, this paper combines the autoregression, differencing, and moving average components of the ARIMA model, along with the application of the SVR model in regression forecasting, optimizing model parameters through the PSO algorithm to minimize forecast errors and improve accuracy. Experimental results show that compared to traditional ARIMA and SVR models, the proposed optimized model has significantly improved prediction accuracy, reflected in lower RMSE, MAE, and MAPE, as well as a higher coefficient of determination (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{R}^{\mathbf{2}}$</tex>). Additionally, through visual analysis, this paper explores the seasonality and intermittency of product demand, as well as potential correlations between different products.