In the modern information network, the personalized recommendation system has become a key part of users in software application. Recommendation algorithms are the core of personalized recommendation systems. Among them, the collaborative filtering is one of the most successful recommendation algorithm in application. However, the traditional collaborative filtering algorithm does not consider user’s multiple interest and measure user’s interest imprecisely, and can’t be applied to recommendation system with kinds of interests. In this paper, a new method of collaborative filtering algorithm based on users’ interest category is proposed using improved fuzzy clustering algorithm to search the nearest neighbors. Finally, the algorithm experiment is given with actual log-data. Results show that the proposed algorithm outperforms the other recommendation ones in efficiency and recommending accuracy.