TIME SERIES FORECASTING AND MODELLING OF FOOD DEMAND SUPPLY CHAIN BASED ON REGRESSORS ANALYSIS

Abstract

Accurate demand forecasting in the food industry is essential due to the perishable nature of many products and the significant waste and financial loss associated with poor inventory management. This project aims to develop a robust time series forecasting model for the food demand supply chain, leveraging advanced machine learning and deep learning techniques. We employed a variety of regression models, including Random Forest Regressor, Gradient Boosting Regressor (GBR), Light Gradient Boosting Machine Regressor (LightGBM), Extreme Gradient Boosting Regressor (XGBoost), Cat Boost Regressor, Long Short-Term Memory (LSTM), Bidirectional LSTM (BiLSTM), and CNN-2D. These models were selected based on their proven effectiveness in handling time series data and their ability to capture complex patterns and dependencies. The models were trained and evaluated using historical sales data from various food products, considering multiple regressor variables such as seasonal trends, promotional activities, and external factors like weather conditions. Our approach integrates feature engineering and hyperparameter tuning to optimize the performance of each model. The results demonstrate that advanced regressors, particularly ensemble methods and deep learning architectures, significantly improve forecasting accuracy, thereby enabling more efficient inventory management and reducing waste. The findings from this study provide valuable insights for stakeholders in the food industry, highlighting the importance of sophisticated forecasting techniques in enhancing supply chain efficiency. By implementing these advanced models, organizations can achieve better alignment between supply and demand, ensuring product availability while minimizing losses due to overstocking or spoilage.