Efficient inventory distribution management is essential to the supply chains’ success. Traditional inventory policies serve as a backbone for inventory management, yet data-driven decision-making can significantly enhance business performance. In this paper, we propose an approach to inventory order management by leveraging machine learning models. To train these models, we generate data through simulation models based on policies related to inventories, including the order-up-to (OUT) policy and its variant. Subsequently, we employ machine learning algorithms including XGBoost, LightGBM, and AdaBoost to optimize inventory management decisions. Our objective is to develop superior inventory policies tailored for members of four-stage serial supply chains, thereby improving overall supply chain performance.