Accurate forecasting of daily demand is vital in contemporary Supply Chain Management (SCM), as it has a direct effect on inventory optimization, cost savings, and overall customer satisfaction. Nevertheless, the unpredictable nature and pronounced seasonal trends in consumer demand data create substantial difficulties in creating predictive models that are both precise and generalizable. This research aims to address these obstacles by developing a robust forecasting framework that utilizes Machine Learning (ML) methods. A thorough analysis of historical daily order data is performed, examining the effectiveness of various ML algorithms, including Linear Regression (LR), Ridge Regression (RR), Random Forest (RF), and XGBoost. In order to improve predictive capabilities, a new Stacked Ensemble (SE) model is introduced, which leverages the advantages of the base learners alongside a Gradient Boosting Regressor (GBR) as the meta-learner. The models are rigorously evaluated using well-known accuracy metrics, such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R2 Score. The findings reveal that the SE model outperforms the others, achieving an R2 of 0.9929, highlighting the effectiveness of ensemble learning in identifying complex demand patterns and providing scalable, highly accurate forecasting solutions for supply chain applications.