Collaborative Ensemble Learning: Combining Collaborative and Content-Based Information Filtering via Hierarchical Bayes

Collaborative filtering (CF) and content-based filtering (CBF) have widely been used information filtering applications, both approaches having their individual strengths and weaknesses. This paper proposes a novel probabilistic framework to unify CF and CBF, named collaborative ensemble learning. Based on content based probabilistic models for each user's preferences (the CBF idea), it combines a society of users' preferences to predict an active user's preferences (the CF idea). While retaining an intuitive explanation, the combination scheme can be interpreted as a hierarchical Bayesian approach in which a common prior distribution is learned from related experiments. It does not require a global training stage and thus can incrementally incorporate new data. We report results based on two data sets, the neuters-21578 text data set and a data base of user opionions on art images. For both data sets, collaborative ensemble achieved excellent performance in terms of recommendation accuracy. In addition to recommendation engines, collaborative ensemble learning is applicable to problems typically solved via classical hierarchical Bayes, like multisensor fusion and multitask learning.