Product Clustering Analysis Based on the Retail Product Knowledge Graph

Abstract

Product clustering analysis is essential in designing retail marketing strategies. It is a common practice that retailers use to effectively manage their product inventory, marketing promotions, etc. The most intuitive way of clustering products is by their explicit attributes, such as brand, size, and flavor. However, these approaches do not integrate the customer-product interactions, thus ignore the implicit product attributes. In this work, we construct a retail product knowledge graph based on Amazon product metadata. Leveraging a state-of-the-art network embedding method, RotatE, our main objective is to unveil hidden interactions of products by including implicit product attributes. These hidden interactions bring insights to downstream operations such as demand forecasting, production planning, assortment optimization, etc.