Research On Collaborative Filtering Recommendation Algorithm Based On Co-similarity

With the rapid development of the Internet, Web has become the main way forpeople to access to information. While e-commerce is widely spreading among people, theissue as to promote users with useful information from the expending information hasbecome a big concern. The emergence of search engine satisfied the need of informationretrieval in a certain degree; however, it cannot meet the users’ demand from differentfields and different levels. Therefore, personalized recommendation technology is bornform the need of information searching, it is one model of personalized service and itsnature is information filtering.In order to overcome information overload, recommender systems have become akey tool for providing users with personalized recommendations on items such as movies,music, books and news. Intrigued by many practical applications, researchers havedeveloped algorithms and systems over the last decade. The classic and widely usedrecommendation method is collaborative filtering. The goal of collaborative filteringsystem is to suggest new items or to predict the utility of items for users based on theusers’ previous likings and the opinions of the other like-minded users. But there still existmany problems, such as the cold-start problem and the data sparsity problem.A collaborative filtering recommendation method based on Co-similarity bringsmethods of social network analysis into collaborative filtering. We propose a collaborativefiltering algorithm based on Co-similarity combines multiple similarity matrices fromunipartite friendship Network, user-item bipartite network and the behavioral networkbased on their behavioral similarities that are computed by using navigational patterns.Moreover, we propose an effective weighting strategy of SRNs influence based on theirstructured density. We perform an experimental comparison of the proposed methodagainst existing rating prediction and product recommendation algorithms using doubandataset. Our experimental results show that the proposed method is more effective.