Research On Collaborative Filtering Recommendation Algorithm Based On Social Network

With the rapid development of Web2.0 technology,the Internet has brought a lot of convenience to people's work and life.For example,we may need to go far shopping malls in order to buy a clothe in the past.But now,we can do this through Taobao,Jingdong or other electronic business website which save our time and effort.However,when mass of goods and information flooded in the Internet,people are faced with a serious problem of information overload.In order to help users find their required items more easily and quickly,recommended system was emergence and is widely applied in actual site,such as Amazon,Netflix and so on.In many recommendation algorithm,collaborative filtering algorithm become one of the most popular and success recommendation algorithm.But it is also faced with the problem of sparse data and poor scalability which affect the effectiveness of the recommendation algorithm.In recent years,with the rapid development of social network,many traditional e-commerce sites also integrate social network to provide comment,communication and feedback ability for users which have produce massive social network information.It provides a new way of thinking to solve the problems of traditional collaborative filtering algorithms.Based on the analysis,we proposed two improved collaborative filtering algorithms by incorporating users social network information.The main content is as follows:(1)In order to solve the problem of data sparsity,we utilize social network information to fill missing values selectively in rating matrix and maximize the use of existing rating information in item-item similarity calculation and user ratings of item prediction period.Then a collaborative filtering recommendation algorithm based on filling of missing values using social network is proposed.Finally,experimental verification is performed on public dataset.Relevant results show that the proposed method has alleviated the data sparseness problem and outperformed other collaborative filtering algorithms in terms of rating error and precision.(2)In order to solve the scalability problem,we present a social community detection algorithm based on user's social network.Then we find item collections corresponding with user community which is divide items into several groups.Next,we can construct a low dimensional and dense item membership matrix instead of high dimensional and sparse rating matrix in the traditional collaborative filtering.Furthermore,we can run item-based collaborative filtering algorithm using the item membership matrix.Finally,experimental verification is performed on public dataset.Relevant results show that we proposed new method greatly improves the computational efficiency while improving the accuracy of the recommendation results compared with other methods.The proposed two algorithms greatly improves the traditional collaborative filtering algorithm processing data sparsity and scalability problems,which makes the recommendation system has better availability and higher accuracy in the real environment applications with growing users and items.