An improved Collaborative Filtering approach based on combined clusters with modified prediction formula
Chu Yang-jie, Xin-wei Chen, Jie Zheng · 2011
This paper describes a new technique for making personalized recommendations. Among existing recommendation algorithms, user-based Collaborative Filtering (CF) approach is the most promising one. However, the problems like multiple-interests lead to a decline in recommendations' quality. To overcome the problem of multiple-interests and dimensionality curse as well, we propose a modified CF method based on combined clusters, which are obtained using EPCH. Further improvements are made by proposing a modified prediction formula. MovieLens dataset is used to evaluate our method and it's demonstrated that the new method outperforms traditional CF approach.