Time Series Forecasting of Food Demand Using Regression-Based Analytical Models in Supply Chains

Abstract

Accurate demand forecasting plays a critical role in the food industry, where products are highly perishable and poor inventory decisions can lead to considerable waste and financial loss. In recent years, machine learning and deep learning methods have shown strong capabilities in capturing patterns within time-dependent data. This study uses the Genpact Food Demand Forecasting dataset to analyze how different factors influence demand and to identify key features that impact order quantities. A comparative analysis is conducted using seven regression models, including Random Forest, Gradient Boosting, LightGBM, XGBoost, CatBoost, Long Short-Term Memory (LSTM), and Bidirectional LSTM. The performance of these models is evaluated to determine their effectiveness in predicting demand. The findings indicate that deep learning approaches, particularly LSTM, provide more accurate forecasts compared to other methods, making them a promising choice for improving decision-making in food supply chain management.