Application of data mining methods for link prediction in social networks

Abstract

Using social networking services is becoming more popular day by day. Social network analysis views relationships in terms of nodes (people) and edges (links or connections—the relationship between the people). The websites of the social networks such as Facebook currently are among the most popular internet services just after giant portals such as Yahoo, MSN and search engines such as Google. One of the main problems in analyzing these networks is the prediction of relationships between people in the network. It is hard to find one method that can identify relation between people in the social network. The purpose of this paper is to forecast the friendship relation between individuals among a social network, especially the likelihood of a relation between an existing member with a new member. For this purpose, we used a few hypotheses to make the graph of relationships between members of social network, and we used the method of logistic regression to complete the graph. Test data from Flickr website are used to evaluate the proposed method. The results show that the method has achieved 99 % accuracy in prediction of friendship relationships.