MACHINE LEARNING APPROACHES FOR ACCURATE DEMAND FORECASTING IN SUPPLY CHAIN MANAGEMENT

Accurate demand forecasting is a cornerstone of effective supply chain management, enabling companies to align production, inventory, and distribution with market needs. Traditional statistical models often fail to capture the nonlinear and complex patterns in consumer demand, particularly in the presence of seasonal shifts, promotional events, and external shocks. In recent years, machine learning (ML) has emerged as a powerful tool for enhancing demand forecasting accuracy by leveraging large-scale historical and real-time data. This paper reviews the core machine learning techniques applied to demand forecasting, including supervised learning, time series forecasting models, and ensemble methods. We develop and evaluate a hybrid forecasting framework that integrates Long Short-Term Memory (LSTM) neural networks with gradient boosting to capture both sequential patterns and feature-based dependencies. The proposed approach is validated using a retail demand dataset, and its performance is benchmarked against traditional models. The results demonstrate that ML-based methods significantly outperform classical forecasting techniques, offering improvements in forecast precision, robustness to noise, and responsiveness to dynamic market signals.