Abstract

Recently, some studies have begun to explore the potential that inventory management combined with machine learning algorithms could provide as a means of producing efficient and flexible inventory management methods. In this way, although there are some methods to carry out this practice, none are set up for impulse purchase products. This article illustrates this perspective within the context of an impulse purchase product provisioning problem and shows how group policies based on a clustering process can result in better (lower cost) groupings. To solve this problem, a method is proposed for finding a near-optimal inventory grouping solution. The key innovation in this solution is the idea to form groups for the items that have similar demand or ordering and cost characteristics. Subsequently, once the clusters have been formed, it was necessary to look at aggregating impulse purchase SKUs, and then two grouping techniques or heuristics that both consider common characteristics and develop some ordering decision rules are presented. The results show that the proposed method can be used to cluster impulse purchase products more effectively and the grouping techniques applied were efficient in terms of solution quality. The aim of the proposed unsupervised clustering-based method was not only to provide a classification of SKUs free of subjectivity processes but also to provide an approach to apply more efficient inventory policies for impulse purchase products.