An Approach of Finding Localized Preferences Based-On Clustering for Collaborative Filtering

Collaborative filtering has been very successful in both research and applications. Current collaborative filtering based on clustering compute the whole set of items during the process of clustering or selecting nearest-neighbors, because the researchers believed if users have similar preferences on some of items, they will have the similar preferences on other items. But we think that users have similar preferences only on parts of items, even they are neighbors and ignoring the fact of traditional methods probably make the prediction result inaccurate. For this reason, we try to propose a new collaborative filtering algorithm by using the localized preferences between users. We design an algorithm based on cluster model to find the localized preferences and then use the localized preferences between users to select neighbors for active users. Experimental results show that our proposed framework can significantly improve the accuracy of predication as well as solve the scalability problem because of the cluster method.