FOOD DEMAND FORECASTING SYSTEM BASED ON MACHINE LEARNING

Tóm tắt

Food demand forecasting plays a crucial role in optimizing the food supply chain, helping to reduce waste and inventory while improving customer service. However, traditional models often struggle to handle nonlinear, noisy, and high-dimensional data in real-world systems. This paper proposes a forecasting system utilizing the XGBoost model, combined with a feature preprocessing pipeline, log transformation, and hyperparameter optimization techniques. On a real-world test dataset, the proposed model achieves high performance, significantly improving prediction accuracy. These results clearly demonstrate the effectiveness of feature engineering and hyperparameter tuning in large-scale food demand forecasting tasks.