On the use of cross-domain user preferences and personality traits in collaborative filtering

Abstract

We present a study comparing collaborative filtering methods enhanced with user personality traits and cross-domain ratings in multiple domains on a relatively large dataset. We show that incorporating additional ratings from source domains allows improving the accuracy of recommendations in a different target domain, and that in certain cases, it is better to enrich user models with both cross-domain ratings and personality trait information.