A New Effective Collaborative Filtering Algorithm Based on User's Interest Partition
He Keqin, Liang He, Xia Weiwei · 2008
Traditional collaborative filtering algorithms all suffer from inaccurate recommendation and bad scalability. In this paper, we propose a new collaborative filtering algorithm based on user?s interest partition. We divides user?s whole interest into pieces. Each piece of interest is called interest unit. And the similarity between users on interest unit is named local similarity. The similarity between users on whole interest is named holistic similarity which is similar with the traditional similarity. Our approach searches the nearest neighbors of active user according to the linear combination of local similarity and holistic similarity. Through experiments, the algorithm can solve the problem of high sparsity on user-item matrix. Our algorithm also has a better quality on recommendation according to experiments.