Research On Collaborative Filtering Recommendation Algorithm Based On Particle Group Clustering

With the thriving era of the Internet, information arises about billions per day. Along with the excessive expansion of the amount of information, "Information Overload" is approaching, as a result, it inconvenient for users to achieve needy information from various information. According to the latest report by the Pew Research Center, "Information Overload" will be a potential problem which must threaten Internet freedom powerfully. At present, filtering mechanism to solve the problem includes the information-retrieval technology and the recommendation system. What we shouldn’t ignore is that the quality of search engine mostly relied on the degrees of the information accuracy described, the more accurate description, the better the search results, besides, search engine can’t deal with the information of pictures and music effectively. However, recommendation system can predict the user’s needs and interests from the users’behaviors (such as behavior score, browsing behavior and add to cart behavior, etc), then it recommend needy information to the users without accurate description. At the same time, recommendation system can deal with the information of pictures and music effectively. It is obvious that recommendation system will be an essential measure to solve the "Information Overload" in the future.Nowadays, the personalized recommendation system has been used widely, for example,e-business, camera and video. As a consequence, a lot of recommendation technologies arise, such as content-based recommendation technology, the collaborative filtering recommendation technology, graph-based technology and so on. What’s more, the collaborative filtering recommendation technology is the most mature technology among them. However, with the abrupt increase of the users and items, the collaborative filtering technology is facing challenges of Cold Start and Data Sparsity.The paper aims mainly at Data Sparsity in user-based collaborative filtering algorithm. The main research is as following:First:Based on the user preference K-means based on collaborative filtering recommendation algorithm, analyzing the clustering result effect on the recommendation accuracy. A collaborative filtering recommendation algorithm based on user preference particle swarm clustering is proposed.The principle and implementation of collaborative filtering recommendation algorithm based on user preference clustering is introduced in detail.Through simulation experiments, it obvious that the particle swarm clustering algorithm can improve the recommendation accuracy in user-based collaborative filtering algorithm when compared to K-means algorithm.Second:Based on the user preference particle swarm clustering based on collaborative filtering recommendation algorithm, analyzing users’choices caused by users’features. A collaborative filtering recommendation algorithm based on user’s comprehensive information particle swarm clustering is proposed. The algorithm considers user features and user preference information to the user for particle swarm clustering and nearest neighbor set selection. Through the simulation experiments, it shows that the algorithm has a higher accuracy than user preference particle swarm clustering based on collaborative filtering recommendation algorithm.Third:Based on user’s comprehensive information particle swarm clustering based on collaborative filtering algorithm, combined with the application of personalized recommendation system in the movie, designed a simple recommendation system. That through collection of user information, recommended movie for users interested.