A Hybrid Machine Learning Framework for Daily Dairy Demand Forecasting: Integrating SARIMAX and XGBoost for Seasonal Production Optimization

Accurate demand forecasting is a critical requirement in the dairy industry, where fluctuations in consumption, seasonal effects, and product perishability create significant operational challenges. This case study systematically compares Extreme Gradient Boosting (XGBoost) and Seasonal AutoRegressive Integrated Moving Average with Exogenous Variables (SARIMAX) for dairy demand forecasting. The models were trained and validated using one year of production and distribution data from a major dairy manufacturing facility in Jordan, representing Middle Eastern market conditions, covering six products: Laban, Shanenah, Labneh, Jameed, Milk, and Zabadi. Novel integration of seasonal decomposition and breakpoint detection techniques identified structural demand shifts during cultural events, particularly Ramadan, providing a methodological framework for incorporating religious and cultural factors into demand forecasting. Forecasting accuracy was evaluated using the Diebold–Mariano test, which confirmed that SARIMAX significantly outperformed XGBoost in most cases (p < 0.05), achieving an average accuracy of 94.27% compared with 92.99%. SARIMAX showed superior performance for highly seasonal products such as Jameed (95.78% vs. 91.46%) and Shanenah (94.41% vs. 93.48%), whereas XGBoost offered marginal advantages for stable-demand products like Milk (94.63% vs. 94.48%). Economic analysis for this facility indicates potential inventory cost reductions of 15–20% and waste decreases of up to 25%, demonstrating the business value of enhanced forecasting accuracy. This case study illustrates the potential for integrating advanced forecasting methods in dairy manufacturing operations, contributing to Industry 4.0 applications in perishable goods supply chains within similar cultural and operational contexts.