Research On Collaborative Filtering Recommendation Algorithm Based On Social Network Information

In recent years,with the development of the network technology,more and more internet companies came into being,and FaceBook.LinkedIn?Tencent as the representative of the social networking sites have appeared.These websites provide people some platforms for learning,communication and entertainment,and greatly enrich people's life and have an important impact on people's way of life.However,most of them have a huge amount of users and massive data information,which is convenient for Internet users to obtain information they want and at the same time brings technical challenges to recommender systems.As one of the most popular recommendation methods,collaborative filtering is widely used in the field of recommendation because of its good scalability and extensibility.The evaluation data provided by users plays a key role in the traditional collaborative filtering algorithm.But in the big data era today,facing massive data information coming from internet,it is difficult for users to evaluate all projects they love,which results in the sparsity of the evaluation data and thereby affects the accuracy of the recommendation algorithm.How to make full use of social network information rich and to establish more reasonable recommendation models to improve the accuracy of recommendation,has become a hot research direction.Aiming at the existing problems in collaborative filtering algorithm,the main research work of this paper is as follows:First,analyzing the characteristics of social network information from different angles.The algorithm idea,implementation steps,advantages and disadvantages of collaborative algorithm are studied in this paper.Starting from the existing problems of collaborative filtering algorithm,we make full use of social network information to improve the algorithm.Second,analyzing the factors which affect the algorithm from several aspects,and studying the similarity model of collaborative filtering algorithm,we propose a collaborative filtering algorithm based on social network information.The similarity models of user(project)play a key role in collaborative filtering.Due to the sparse data,the similarity model based on the evaluation data can not improve the recommendation accuracy.Aiming at the above problems,in tthis paper we use the social network information to establish the user similarity model from three aspects:the user's information,the label information and the social relationship,and then use collaborative filtering algorithm.Third,the nearest neighbor of collaborative filtering algorithm is studied.The traditional collaborative filtering algorithms rely on its nearest neighbors,and the accuracy of the nearest neighbor has an important influence on the accuracy of the recommendation.However,the nearest neighbor acquisition often depends on the similarity model.In this paper,we propose a collaborative filtering algorithm improved nearest neighbor by researching on the social network information and the nearest neighbor.In order to search for the nearest neighbor of the user and the project,the new algorithm establishes the similarity models from the evaluation data,the user's social activity information and the project text information.Then the obtained four nearest neighbor are combined to generate the denoising nearest neighbor and the supplement nearest neighbor.Finally,we use the two nearest neighbors improved to combine the traditional collaborative filtering algorithm and have done some experiments with KDD CUP 2012 Track1 data sets,which results show that the proposed two algorithms have effectively alleviate the data sparseness problem,compared with the traditional collaborative filtering algorithm,the recommendation accuracy is improved.