Improving Recommendations in Tag-based Systems with Spectral Clustering of Tag Neighbors

Tag as a useful metadata reflects the collaborative and conceptual features of documents in social collaborative annotation systems. In this paper, we propose a collaborative approach for expanding tag neighbors and investigate the spectral clustering algorithm to filter out noisy tag neighbors in order to get appropriate recommendation for users. The preliminary experiments have been conducted on MovieLens dataset to compare our proposed approach with the traditional collaborative filtering recommendation approach and naive tag neighbors expansion approach in terms of precision, and the result demonstrates that our approach could considerably improve the performance of recommendations.