Research And Implementation Of Recommendation Based On User's Social Networks Information

With the rapid development of the Internet and information technology, personalized recommendation technology has been widely studied and applied as an important information filtering solutions, and the major Internet sites have provided personalized recommendation service to enhance the core competitiveness of the site. Although collaborative filtering algorithm has been widely used in recommender systems and get a great success, the network resource information and the number of users get a dramatic increase with the development of the internet, which makes collaborative filtering development face enormous challenges, such as data sparsity and cold start problem. In recent years, the popularity of social media represented by FaceBook, micro-blog, WeChat provides a new idea for personalized recommendation research. By analyzing and mining user's mass of information and behavior and social relationship in the social networks, we can know his interests, then provide more accurate personalized service for him. In order to alleviate the problem that data sparseness leads to poor recommendation quality, this paper will combine the user's social network relationship with collaborative filtering recommendation algorithm to improve the accuracy of the recommendation. The specific research work of this paper are as follows:(1)Study and analyze collaborative filtering algorithm and its disadvantage in detail. For the problem of data sparseness, it proposes a combination recommendation algorithm. The idea of the algorithm is that combine the Slope One algorithm with user-based collaborative filtering algorithm by means of superposition. Using the advantages of simplicity and high efficiency and accuracy of Slope One to fill score matrix to alleviate the problem of sparse data, finally the detailed realization of the algorithm are given.(2)Analyze several classical similarity algorithm and shortcoming of recommendation system. To overcome the disadvantages, it analyze the significance of social network information for recommendation. Using personal preference and social network relationship characteristic make recommendations. By combining the rating similarity calculated by user preference information and the social similarity calculated by social networking relationships, it not only can improve the recommendation quality, but also increase reliability of the recommendation system. To further enhance the accuracy, it uses the coefficient based on the number of common-Rated items to correct the of similarity model. Finally, it introduces the recommended process of collaborative filtering recommendation algorithm.(3)Realize the algorithm using Mahout, and use Movielens dataset and Baidu recommend competition dataset to analyze and verify the algorithm. Compared the improved algorithm with the traditional collaborative filtering algorithm, verifies the feasibility of the improved scheme.