Genetic Algorithms for Discovering Community Cores Hidden within Multidimensional Social Networks

Abstract

Multidimensionality in social networks is a significant problem that appeared owing to the fact that most social networking sites such as YouTube, Twitter, and Facebook enable people to interact with each other through various social activities, which reflects various types of relationships between them. Each type of these relationships is represented by a network dimension. In resent years, discovering community cores hidden within multidimensional social networks has become essential in social network analysis field. When dealing with this type of networks, the definition of community detection problem changes to be, the discovery of the shared community cores across all network dimensions. Many studies presented on the topic of community detection, but communities uncovering in multidimensional networks has not deeply investigated yet. In this paper algorithms to solve community detection problem are proposed as a single-objective and as a multi-objective optimization problem by applying genetic algorithms. The most popular objective functions proposed over the past years are used. The performance, the limitation of each objective, and how those objectives correlate with each other has been clarified. Experiments in synthetic and real world networks showed the ability of the proposed algorithms to correctly discover community structures that maximize modularity across all network dimensions, compared with others on the literature. The proposed algorithms have the advantages that it doesn’t need any prior knowledge about number of the hidden community cores.