A Generic Framework for Cross Domain Recommendation

Abstract

Cross-domain recommender systems research focuses on recommendation improvement in the target domain based on knowledge gathered from the source domain. Recently, a number of review studies were published, however, they did not provide their reader with a step-by-step guide to conduct cross-domain recommender systems research. In this paper, we present a generic cross domain recommender systems framework which identified different phases and components required to execute a cross-domain recommendation research. The proposed framework is based on analysis of primary studies collected from recently published high impact review studies. It was found that all of the collected primary studies included identified phases, however, each component had multiple options (approaches) which can be applied based on scenarios. The proposed framework can be used for analysis and comparison of existing primary studies, and also explains why cross domain recommender systems approaches are evaluated the same as conventional recommender system approaches. Finally, conclusion and future direction are presented.