K.Teja ,J.Vamsi 03 06 Demand forecasting is a fundamental component in supply chain management enabling organizations to predict future product demand and optimize inventory production and logistics Traditional statistical approaches often struggle with nonlinear relationships seasonality and highdimensional data This paper proposes an advanced forecasting framework integrating machine learning and deep learning modelsRandom Forest RF Gradient Boosting GB and Long ShortTerm Memory LSTM The study emphasizes handling temporal dependencies feature interactions and uncertainty in demand patterns A comparative analysis is conducted using realworldlike datasets evaluating models based on MAE RMSE and MAPE Experimental findings indicate that LSTM captures temporal dependencies effectively while ensemble methods ensure robustness and interpretability The hybrid modeling approach significantly improves forecasting accuracy and scalability in dynamic environments Copyright (c) Journal of Science Engineering Technology and Management Science. All rights reserved 2026