Research And Implementation Of The Personalized Recommendation System Based On SNS

With the rapid development of information technology, the problem of ‘information overload’ becomes worse and worse. Personalized recommendation, as the important way to solve the problem of information overload has been hotspot which many scholars and companies research on. However, the existed personalized recommendation systems still have problems such as cold start, data sparse and hybrid of different recommendation algorithms. The new method of recommendation will be studied.As the existed personalization recommendation systems have little study in the structure of networks, to consider the important of information propagation in the social networking, and analyze the SNS, this paper presents a personalized recommendation algorithms based on social networking structure, and builds the recommendation system with data of micro blogs. In this paper, the main research works include:1. We research the basic structure of SNS. Aiming at the Hierarchical Clustering Algorithm which doesn’t have certain condition to end and doesn’t know how many communities to divide, and the classic GN algorithm has high complexity, we present an active node-based algorithm of community discovery. The algorithm chooses some active node as the centers of clustering, and clusters the other nodes by the distances from active nodes. Then, we implement the algorithm with MapReduce framework to improve the efficiency of execution.2. We pay attention to a linked-based similarity computing method called SimRank and its parallel implementation. Then we analyze the improved algorithm of Delta-SimRank. Aiming at the problem that the parallel Delta-SimRank doesn’t consider the directions of edges at Map step, we improve the MapReduce implementation of Detlta-SimRank. At last, we confirm the improved algorithm with Facebook Dataset.3. Combine with the community structure and SimRank algorithm, we present a personalized recommendation algorithm based on the structure of networking. Then we design a personalized recommendation system based on SNS with Hadoop cloud computing platform. Finally, we use the data of Sina micro blogs to check the availability of the system.