Demand forecasting remains a critical aspect of supply chain management, playing a pivotal role in business planning and resource optimization. Traditionally, forecasting relied on limited data sources, but today, technological advancements and data analytics have revolutionized the process. Machine learning algorithms such as CatBoost, Random Forest, Support Vector Regression (SVR), and XGBoost have emerged as superior alternatives to traditional statistical methods. These algorithms excel in capturing complex demand patterns, customer preferences, and feedback, thus enhancing the accuracy of predictions. This research study aims to showcase the efficiency and robustness of boosting algorithms. By integrating CatBoost alongside other established methods like Random Forest, SVR, and XGBoost, we seek to demonstrate their collective capability in forecasting rice sales demand accurately and efficiently. This research study highlights the novelty of incorporating Extreme Learning Machines into this combination, their potential to enhance predictive accuracy and scalability in demand forecasting tasks.