Research On Collaborative Filtering Recommendation Algorithms For Data Sparsity

With the widespread Internet and development of E-commerce, Web has become an important way to access information. But on the other side, with the increasing web information, people have to spend much time on finding the interesting content we just need. Personalized recommendation has emerged in response to the information overload problem. Collaborative filtering (CF) is one of the most successful technologies in recommender systems, and widely used in many personalized recommendation systems. However, most collaborative filtering algorithms suffer from data sparsity which leads to inaccuracy of recommendation.Data sparsity is the inevitable result with increasing of the number of user and item. And, Collaborative filtering algorithm is based on the user's historical ratings. Therefore, data sparsity is an important factor which constraints the result of collaborative filtering algorithm in accuracy.This paper mainly focuses on data sparsity of collaborative filtering algorithms, and analyses data sparsity on collaborative filtering algorithms from two points. Based on this, two improved algorithms are proposed in this paper. One is A Collaborative Filtering Algorithm Combining the Feature of Controversy, and the other is A Collaborative Filtering Algorithm based on item recursive. Finally, we apply the proposed algorithms to the library interaction system for education and research to test its practicability.The main contributions of this paper are as follows:1) Proposed a new CF algorithm, which could reduce the inaccurate similarity in data sparsity. We consider the whole ratings between items and propose the conception of "Item Controversy Similarity(ICS)",which measures the items'similarity by calculating the divergence of variance of the rating values between items. Combing the ICS to the traditional similarity calculation algorithm, we propose a new CF algorithm, which could reduce the inaccurate similarity in data sparsity. Empirical studies on dataset MovieLens show that algorithm outperforms other state-of-the-art CF algorithms and it is more robust against data sparsity.2) Proposed an improved CF algorithm to against data sparsity. We analyze the reasons why data sparsity become more serious when traditional CF algorithms predict the rating of the target user for the target item. Based on this, we utilize recurrence relation of items to avoid data sparsity deterioration. The experimental results show that the proposed method can improve recommendation performance.3) Finally, we apply the proposed algorithms to the library interaction system for education and research to test its practicability.