Hybrid Approach for Predicting and Recommending Links in Social Networks

Abstract

The prevalence of social networks like Myspace, Facebook, Hi5, and LinkedIn is increasing day by day, as they provide a platform where users of the social network can share their content. They also facilitate their users by recommending new friends on the basis of local or global network features. Local feature-based approaches do not exploit the whole network structure. In contrary to the techniques based on local features, global feature-based techniques make use of the complete network structure, being less efficient for large social networks. Here, we define a hybrid feature-based approach that uses local graph feature by computing proximity between every pair of nodes. It also captures global feature by computing all length two and length three pathways between each pair of vertices of the network. We performed experimental evaluation by comparing the proposed approach with other friend recommendation techniques. The experimental results indicate that our algorithm provides adequate level of efficiency as well as accuracy in friend recommendations.