With the rise of YouTube, Flickr, Last.fm and other social networks, tagging systems play an increasingly important role in our everyday life. Analyzing user's tagging behavior of different resources is very important in providing high quality services. In this paper, major community structure detection algorithms are implemented in clustering analysis in tagging system. By comparing their performances on different datasets, a personalized recommendation algorithm for tagging system is designed. Experimental results indicate that the proposed algorithm performs well in detecting different user interests and thus enhances the quality of the recommendation system..