Global challenges of food insufficiency and food wastage remain prevalent. Artificial intelligence provides the technology needed for supply chain systems to minimize food waste. In this context, an efficient LSTM-XGBoost hybrid model is proposed here to enhance demand forecasting in supply chain optimization and reduce food waste. In the proposed work, the LSTM networks manage temporal dependencies from the prepared dataset in the first stage. The second phase incorporates XGBoost to utilize the extracted features from LSTM, identifying nonlinear relationships and residual patterns that LSTM might miss on its own. The proposed hybrid method guarantees computational efficiency by transforming high-dimensional outputs of LSTM into significant features, resulting in decreased complexity for XG-Boost while preserving our essential temporal data. The model was evaluated using a real-world food supply dataset and achieved an R \(^2\) score of 0.91 and RMSE of 116.60, outperforming several baseline models, including standalone LSTM and XGBoost. The result confirms the effectiveness of the proposed model over other existing models. In addition, the proposed model incorporates external factors such as holidays and seasonal variations, enabling it to more effectively anticipate demand fluctuations and irregularitiesqueryPlease check and confirm if the authors given and family names have been correctly identified..