Enhancing Material Backorder Prediction in Supply Chain Management: A Comparative Study of Tree-Based Models

Material backorder prediction is a pivotal problem in supply chain studies that affects an inventory system’s effectiveness and service level. Identifying parts with the greatest chance of unavailability before they occur can be a significant opportunity to improve the performance of an organization. Accurate backorder prediction is critical for producers with limited production capacity. A more advanced prediction method can help predict future orders. In this study, we used five different tree-based models: Random Forest Classifier, Decision Tree Classifier, and Gradient Boosting Classifier, KNeighbors, Adaboosting models. With the help of Spearman feature extraction method, and random over sampler, we found the optimal combination of their hyperparameters. We applied two sampling methods to deal with imbalanced data to improve the accuracy of the output in any classification problem. Accordingly, the result of the proposed method showed an accuracy of ۹۹.۲۹%, which indicates an almost perfect classification model.

کلیدواژه ها:

Machine Learning ، Classifiers ، Random Forest ، Gradient Boosting ، Decision Tree ، KNeighbors ، Adaboost ، Supply Chain ، Predication ، Back Order Problem.