Collaborative Filtering Recommendation Algorithm Based On Expert Users

Due to the rapid development of the Internet in recent years, allowing users to have to deal with huge amounts of data, information overload is serious. The emergence and development of recommender systems effectively help users decision-making process. How to generate a new recommendation algorithm or the improvement of existing recommendation algorithms mainstream of current research in the field of recommender systems. However, with regard to different types of users on the recommender systems were much influence in-depth study.First, We assume the existence of a part of the expert users and expert users can produce reliable recommendation. Traditional collaborative filtering recommender systems is the use of purchase history data of users with similar interests, recommended items for the current user. But traditional collaborative filtering recommender systems is still not effectively solve the cold start, the data sparse and difficult to calculate, such as high-dimensional data. Studies have shown that, compared to traditional users get similar recommendation, people tend to a friend’s recommendation, preference similarity and trust between people with a positive relationship.Then, in order to generate recommendations for target users more effectively, we identified a similar expert neighbor similarity calculation by the target user and expert users to generate recommendations, rather than the target by finding similar users neighbor set of ordinary users to generate recommendations. As the amount of data to be processed increases, the recommender systems except to solve the cold start and sparsity two traditional problems, but also have to solve the system’s scalability problems, ensure that the system can accept user recommendation results within the given time at this stage. The nearest neighbor recommended collaborative filtering algorithms can produce very good results. However, the conventional nearest neighbor based collaborative filtering algorithm is still exit the above-mentioned problems. In this paper, we improved conventional nearest neighbor collaborative filtering algorithm, and we propose a collaborative filtering recommendation algorithm based on expert users.Furthermore, in order to improve the performance of our recommendation only use a small portion of user-generated ratings of experts recommendation, we also proposed a new method to determine the concentration of expert users from all user ratings data.Finally, we explore this new approach to solve many of the problems of traditional systems filter capabilities.