Explainable product backorder prediction exploiting CNN: Introducing explainable models in businesses

Abstract Due to expected positive impacts on business, the application of artificial intelligence has been widely increased. The decision-making procedures those models are often complex and not easily understandable company’s stakeholders, i.e. people having follow up recommendations or try understand automated decisions a system. This opaqueness black-box nature might hinder adoption, as users struggle make sense trust predictions AI models. Recent research eXplainable Artificial Intelligence (XAI) focused mainly explaining experts with purpose debugging improving performance In this article, we explore how such systems could be made explainable stakeholders. For doing so, propose new convolutional neural network (CNN)-based predictive model for product backorder prediction in inventory management. Backorders orders that customers place products currently stock. company now takes risk produce acquire backordered while meantime, can cancel their if too long, leaving unsold items inventory. Hence, strategic management, companies need based assumptions. Our argument is these tasks improved by offering explanations recommendations. our investigates provided, employing Shapley additive explain overall models’ priority decision-making. Besides that, introduce locally interpretable surrogate any individual model. experimental results demonstrate effectiveness predicting backorders terms standard evaluation metrics outperform known related works AUC 0.9489. approach demonstrates current limitations technologies addressed business domain.