This paper presents a hybrid framework to improve food demand forecasting and reduce food waste in supply chain management. Traditional forecasting methods lack to address the uncertainties in food demand, leading to significant food loss. By integrating the autoregressive integrated moving average (ARIMA) model for capturing linear trends with a Deep Q-Network (DQN) for adaptive decision-making in the proposed work, the proposed hybrid model can overcome these limitations. Utilizing a comprehensive dataset, the ARIMA model effectively captures seasonal trends, while the DQN enhances prediction accuracy by learning from real-time fluctuations and market conditions. The results demonstrate that the proposed hybrid model has achieved a mean absolute error (MAE) of 0.123 and an R-squared (R2) value of 0.92, significantly outperforming the standalone ARIMA model (MAE of 0.150, R-squared of 0.80) DQN model (MAE of 0.130, R-squared of 0.85) and XGBoost Model (MAE of 0.136, R-squared of 0.866). The findings highlight the efficacy of hybridization, presenting a promising solution for enhancing food demand forecasting and addressing the critical issue of food waste within global supply chains.