Context and Customer Behavior in Recommendation

The last few years have seen an increased interest in incorporating context within recommender systems. However, little empirical evidence has emerged to support the premise that context can actually improve recommendation accuracy. Indeed little agreement exists as to what represents the context of a user or indeed how such context should be used within a recommendation strategy. In this paper we study the effect of incorporating contextual variables, both observable and derived from past user behavior, on the accuracy of a content based recommender system. The system was evaluated using data from an Italian online retailer. Results suggest a significant improvement in performance when using contextual variables.