A Graph-Based Recommender for Enhancing the Assortment of Web Shops

Abstract

In this work, we consider a situation where multiple Providers (competitors) serve a common market, using a common infrastructure of sales channels. More speci cally, we focus on multiple web shops that are run by the same web shop platform provider. Our goal is to recommend new items to complement the assortment of a provider, based on user behaviour in the other shops of the same platform. For this new problem, we propose to capture information on how items sell together in a shared product co-occurrence graph. We then adapt known graph-based recommenders to the problem. Further criteria for ranking recommended items are derived as part of a case study conducted in the context of IT web shops. They are combined with the scores of the graph recommenders in a nal ranking function. We evaluate this function with data from our case study context and based on judgments of one shop owner. Our results show that a good ranking can be achieved, reflecting the needs of the shop owner.