Fusing SVR with PSO Improves E-commerce Sales Prediction with 8.98% MAPE

Abstract

This study explores the integration of Support Vector Regression (SVR) with Particle Swarm Optimization (PSO) to forecast clothing product sales at Nara Gallery Collection Boutique, addressing the challenge of achieving high forecast accuracy in e-commerce. Through literature review, direct observation, and interviews with a textile SME owner, SVR parameters are optimized using PSO. Results indicate a Mean Absolute Percentage Error (MAPE) value of 8.98% with optimized parameters (C = 34.3642, ε = 0.0110, σ = 0.3677, cLR = 0.1062, λ = 0.0117), enhancing decision-making in inventory management and strategic planning for e-commerce businesses. This research highlights the potential of integrating SVR with PSO for accurate sales forecasting and suggests avenues for further exploration in alternative forecasting methods and optimization techniques. Highlight: Enhanced Forecasting Accuracy: SVR and PSO integration improves e-commerce sales predictions. Parameter Optimization: PSO optimizes SVR parameters, reducing Mean Absolute Percentage Error. Strategic Inventory Management: Accurate forecasts aid in effective e-commerce inventory control. Keywoard: Support Vector Regression, Particle Swarm Optimization, Sales Forecasting, E-commerce, Inventory Management