Community mining in dynamic social networks — Clustering based improved ant colony algorithm

Community mining has been the focus of many recent researches on dynamic social networks. In this paper, we propose a clustering based improved ant colony algorithm (CIACA) for community mining in social networks. The CIACA combines the local pheromone update rule with the global update rule and utilizes heuristic function to adjust the clustering solution dynamically, assisted by decay coefficient of dynamic network model. In order to improve clustering accuracy and convergence rate in the process of ant migration, a structure tightness between nodes based clustering centers initializing method is proposed, which can provide us initial clustering centers with certain clustering precision and high diversity. In addition, random number and specific parameter are used in the ant transition probability, which strengthens the search stochastic properties of CIACA effectively. The proposed CIACA is tested on some benchmark social networks, and is compared with current representative algorithms in community mining. Experimental results show the feasibility and validity of CIACA.