Graph-based Friend Recommendation in Social Networks using Artificial Bee Colony

Friend recommendation is a fundamental problem in online social networks, which aims to recommend new links for each user. In this paper, a new methodology based on graph topology and artificial bee colony is proposed to effective friend recommendation in social networks. In proposed approach, a sub-graph of network is composed by the study user and all the other connected users separately by three degree of separation from the root user. The proposed recommendation system computes four parameters within the generated sub-graph, and suggests the new links for the root user. Artificial bee colony is applied to optimize the relative importance of the weights of each parameter. To verify the proposed methodology, we chose a graph with 1000 members from YouTube. We considered the 20% of all links within the network graph to learning the system using artificial bee colony algorithm. These links were removed from the graph, and a data was generated by using all candidate nodes within the resulted graph, to be a recommend. Then, the generated data were divided into training set and evaluation set. Obtained results demonstrated the robustness of proposed approach with a 36% return rate.