Research of collaborative filtering algorithm based on the probabilistic clustering model

Recommendation system can reduce the information overload and push the right information to the right people in suitable time at suitable occasion. Classical collaborative filtering (CF) approaches are memory based, which recommend the neighbor's favorite information to the users. The method is time-consuming and can't compute the recommending value of every user to every information item. This paper introduces a novel approach based on the probabilistic clustering model to solve the problems. In the approach, we assume that the users and information items can be clustered into different USER Models and ITEM Models with one probability. We can compute the rating of each USER Model to each Item Model. A comparative evaluation of the algorithm and a well-established baseline method on the benchmark datasets shows that: our algorithm can compute the recommending value of every user to every item in a shorter time, and the effectiveness is competitive to other recommending approaches.