Inventory management remains a critical determinant of supply chain performance, directly influencing cost structures, customer satisfaction, and operational efficiency across industries. At a broad level, traditional inventory control approaches such as Economic Order Quantity (EOQ) and rule-based replenishment systems often rely on static assumptions and limited data inputs, making them inadequate for handling demand volatility, supply disruptions, and complex multi-echelon networks. The emergence of data-driven methodologies, powered by advanced analytics and machine learning, provides new opportunities to optimize inventory decisions through real-time insights and predictive capabilities. Narrowing the focus, this study develops a data-driven optimization framework that integrates demand forecasting, cost modeling, and service level constraints to enhance inventory performance. Using historical demand patterns, lead time variability, and operational data, machine learning models are employed to generate accurate forecasts, while optimization algorithms dynamically determine reorder points and order quantities. The framework explicitly balances holding, ordering, and shortage costs against service level targets, enabling adaptive decision-making under uncertainty. Empirical results demonstrate significant reductions in total inventory costs, improved fill rates, and enhanced operational responsiveness. The findings highlight that the integration of predictive analytics with optimization techniques offers a robust and scalable solution for achieving cost efficiency while maintaining high service levels in modern inventory systems.