An explainable machine learning pipeline for backorder prediction in inventory management systems

Abstract

Backorders occur when a product is out of stock, but the costumer is willing to place an order for this product and wait until it will be available for shipment instead of purchasing another. It is an important part of the inventory system contributing to the total costs of the production. Hence, it is important for companies to be able to predict when a product will be backordered to develop mitigation strategies and reorganize their production. Limited studies have focused on the prediction of backorders, a high imbalanced binary classification problem that needs special treatment. However, no previous study has aimed to explain and interpret the main features that contribute to the prediction task. To this end, in this study a machine learning pipeline is developed supported by an explainability analysis in order to identify the most important features that contribute to the prediction of backorders. The results showed that the inventory level of a product combined with the forecast demands and transit time play are the main factors that lead to products' backordering. © 2021 ACM.