An Adaptive Personalized Recommendation Strategy Featuring Context Sensitive Content Adaptation

In this paper, we present a new approach that is a synergy of itembased Collaborative Filtering (CF) and Case Based Reasoning (CBR) for personalized recommendations. We present a two-phase strategy: in phase I, we developed a context-sensitive item-based CF method that leverages the original past recommendations of peers via ratings performed on various information items. In phase II, we further personalize the information items comprising multiple components using a CBR-based compositional adaptation technique to selectively collect the most relevant information components and combine them into one composite recommendation. In this way, our approach allows finegrained information filtering by operating at the constituent elements of an information item as opposed to the entire information item. We show that our strategy improves the quality and relevancy of the recommendations in terms of its appropriateness to the user’s needs and interests, and validated by statistical significance tests. We demonstrate the working of our strategy by recommending personalized music playlists.