A Data-Driven Framework for Profit Modeling and Forecasting in High-Purity Silicon Wafer Production

Abstract

The high-purity silicon wafer industry faces significant challenges due to high energy consumption and complex multi-variable interactions, necessitating precise tools for optimizing production and sales strategies. This study proposes a data-driven framework integrating a monthly profit calculation model with multi-modal forecasting techniques to enhance operational efficiency. A profit model was developed incorporating sales revenue, variable costs, fixed costs, and tax constraints, achieving a net profit prediction error of less than 0.8%. Sensitivity analysis reveals that a 1% increase in sales price boosts profits by 4.5%, while a 5% reduction in monocrystalline silicon ingot prices significantly lowers variable costs. Additionally, a hybrid forecasting approach combining Holt-Winters, LSTM, ARIMA, and GARCH models predicts key factors such as sales volume, sales price, and raw material costs with an average accuracy of 92%. This framework provides actionable insights for dynamic pricing and production planning, offering a replicable paradigm for high-volatility industries. The study contributes to industrial decision-making by integrating advanced forecasting techniques, with practical implications for cost control and market competitiveness.