The Research On The Relation Predicting And User Influence Evaluating Algorithm For Social Network

In recent years, with the rapid development of Internet technology and mobile communication technology, social network service becomes the emerging network application. The social relation prediction and the influence evaluation of social network users are research hotspots in the field. By the deep web information mining, we can analyze the social links that the network will produce. We can also draw the social influence measurement of the users. Then the results will contribute to friends providing, personalized service and guiding the network public opinion. So the research has important social and economic benefits.This paper firstly studies the method for predicting social relation in a social network. Considering social network’s complex structure and large information amount, the probability model and the maximum likelihood model are not suitable for the requirements of social analysis because they cost too much calculation resource. This paper proposes a social link prediction algorithm based on weighted neighborship (WN algorithm). Based on the local structure information, the neighbors of the common neighbors are divided into three classes according to their contacts and are given different weights. In several social network datasets, using the AUC evaluation index, this algorithm is compared to other methods to verify the accuracy of the proposed algorithm.This paper also studies the user influence in the social network. The paper first analyzes the calculation process of the PageRank algorithm. Then, the paper proposes a user influence evaluating algorithm based on composite relation network (CRN algorithm). This paper defines the influence as the two-dimensional vector which consists of diffusion degree and recognition degree. This paper proposes a method for evaluating the reasonableness of influence measurement using the accuracy of link prediction. This algorithm is compared with the original PageRank algorithm using the Spearman and Kendall rank correlation coefficient, verifying the rationality of influence ranking.This paper finally studies the connection between social relation prediction and user influence in social network, combining two parts. This paper puts forward a social relation predicting algorithm based on the influence preferential attachment (IPA algorithm). This algorithm uses the product of the influence measurement and the common neighbor number as a weighted association degree. In several social network datasets, using the AUC evaluation index, this algorithm is compared to other methods to verify the accuracy of the proposed algorithm.