Abstract: Web2. 0 tag recommender systems always contain a lot of redundant special label. Redundancy on tags may even burden the user with additional effort by selecting their preferred items, and these redundancy can introduce erroneous features into the user profile and hamper the effort to judge recommendations. There is usually a lot of irrelevant or redundant features in high-dimensional data sets, feature subsets between different clusters are not the same. Therefore,we should focus on the different features subsets to discover the cluster. This paper proposed subspace PSO clustering to identify tag redundancy. We designed a suitable weighting K-means o均ective function, which is more sensitive to the change variables in weight. On this basis we developed PSO to optimize the objective function, then obtain global optimal value, and finally improve tag redundancy accuracy. Our experimental results show that the proposed algorithm has greater searching capability and obtains a better clustering accuracy.

Key words: Web2. 0 tag recommender systems, Tag redundancy,Subspace PSO clustering