Research On Collaborative Filtering And Content Based Hybrid Recommendation

The rapid development of Internet technology has changed people’s live style essentially. Shopping on e-commerce websites is becoming an important choice for us. However, despite of the convenience, users, people have to face the problem of information overload because of the huge quantity of item information. Personalized recommendation systems came into being just at this moment. Recommendation systems analyze the relationship between users and items and mine users’behavior to find there interest, helping users to choose their favorite ones from all the items. Hybrid recommendation system is the most common form, and the different recommendation algorithms used by it are its core. The main problem of the current recommendation system is the recommendation efficiency of its algorithms.In this paper, I introduce the research background and current development of recommendation systems, and give the illustration of common recommendation algorithms and related technologies. Then I focus on the collaborative-based filtering and content-based filtering algorithms, and improve them respectively. At last I give the discussion of hybrid recommendation prototype system.What is accomplished in this paper is as follows:1) I propose the improved algorithm of Slope One algorithm. Due to the lack of correlation for the items and the adaptability of sparsity, this paper proposes an improved Slope One algorithm using the integration of item similarity with singular value decomposition and verifies its accuracy and improvement of sparsity experimentally.2) I propose the improved algorithm of the vague set content-based recommendation algorithm. To solve the problems of inefficient recommendation and redundant computation, this paper uses the k-means clustering of the vague set to narrow the scope of the final recommendation. According to the experiment, this new algorithm is verified to enhance the efficiency of recommendation.3) I design the hybrid scheme of the two algorithms and the hybrid recommendation prototype system for movies. I firstly propose the hybrid scheme of the above-mentioned two improved algorithms and then give the requirements analysis, architecture, functional structure, database structure, main processes and finally realize the whole system.