Influence-based approach to market basket analysis

Abstract

In this article, we propose an approach to market basket analysis based on the notion of social influence. While traditional market basket analysis looks for combinations of products that frequently co-occur in transactions, we seek to find a set of influential products that, if bought by a customer, will increase the sales volume of the shop. We believe that customers who purchase influential products would also be influenced to purchase other products. We validated our approach with two real-world datasets collected from online shoppings and one dataset collected from a supermarket concluding that influential products identified by our approach increase the influence spread with respect to different baselines: best-selling, highest centrality, frequent sequence initiator, and most promoted products.