The main objective of a supply chain is to fulfil the customer demands in the right quantity at the right time. If an organization cannot meet the customer demand, stockout or out-of-stock situations happen. Inventory stockouts are costly to the organization and are very common in supply chains. It not only results in the loss of revenue but also affects the service level, which results in a loss of competitive advantage. Stockouts can be prevented by proper inventory planning and control. For any organization, it is crucial to maintain optimum inventory levels to ensure customer satisfaction. These days, technologies like Artificial Intelligence and Machine Learning give organizations the ability to foresee the future and proactively manage their inventory in the supply chain. This study investigates various supervised machine learning classifiers. It proposes the best machine learning stock out prediction model for each member of a four-stage divergent supply chain with eight members. Since the dataset for such a supply chain is not available, a supply chain operation simulation has been conducted under three distinct inventory replenishing policies such as Order-Up-To (OUT), Order-Up-To Smoothing (OUTS) and (s, S). Data generated from the simulation is divided into two sets, the train set and the test set. Various supervised machine learning algorithms are trained using training dataset. A five-fold cross-validation technique is adopted for the validation of the model. A random search cross-validation technique adjusts the hyperparameters of each classifier. The performance of the model is evaluated on the test dataset based on the assessment matrices, and it is found that boosting algorithms perform better than the other classifiers. This study proposes a meta-learning-based stacked ensemble model, using XG boost, Ada boost and random forest classifiers as the base model. The performance evaluation shows that for each supply chain member, the stacked ensemble model performs better than other classifiers, including boosting classifiers.