A hybrid recommendation approach based on social tagging data preprocession

As an important explicit rating approach, social tagging can not only describe resources but also reflect user's preferences. Therefore personalized recommendation based on social tagging has becoming a hot research direction. However, recommendation algorithms based on tags will encounter great data sparseness problem. In this paper, we process the original dataset by applying similarity propagation algorithm and popularity dimensionality reduction techniques. Hence the sparseness problem of the dataset can be partially solved. Finally, based on the high-quality dataset, we propose a hybrid recommendation algorithm. The experimental results show that our algorithm has a better performance than traditional pure content based or collaborative filtering recommendation algorithms.