Community Detection In Signed Social Networks Using Conical Area Evolutionary Algorithm

In recent years,mining and analysis of community structure in social networks has been received more and more attention from all of field.In social networks,community structure is one of the most common and important topologies.Therefore,how to find community structures in the social network is of great significance.In real social networks,there are both positive connections and negative connections between nodes.Those social networks,which have both positive links and negative links are called signed social networks.However,most of current community detection method are designed for unsigned networks.Therefore,it is significant to explore and design effective and efficient community detection methods for signed social networks.In this paper,we combine the community detection problem for signed networks with a multi-objective evolutionary algorithm,which makes full use of the intrinsic properties of the bi-objective in community detection and the excellent performance of the conical area evolutionary algorithm.Finally,two community detection methods for signed networks based on conical area evolutionary algorithm are researched and designed in this paper.The main contributions of this paper includes:1.According to the advantages of conical area evolutionary algorithm in the bi-objective optimization and the characteristics of the signed network structure,a crossover operator and a mutation operator,both of which are applicable to signed networks,are designed,as well as a bi-objective model based on weight is established.Finally,a conical area evolutionary algorithm based on weight for community detection in signed networks is proposed and designed,which is also called as CAEAw-SN.2.By introducing and decomposing the modularity indicator Q-modularity,a bi-objective model based on the Q-modularity is established,as well as a new tournament selection mechanism based on Q-modularity is designed,Therefore,a conical area evolutionary algorithm based on Q-modurality for community detection in signed networks is proposed and designed,which is also called CAEAq-SN.3.Firstly,four benchmark signed networks consisting of two illustrated signed networks and two real signed networks are used to test and evaluate the performance of the two proposed algorithms;And then a random signed network generator with parameters is designed,which can not only control the scale of the random signed network,but also control the noise level of community structures by adjusting the value of each parameter.Finally,four groups of random signed networks with normal parameters and 180 random signed networks which are obtained by adjusting the generated parameters ?,+,-respectively from two of them are used to further test and evaluate the performance of the two proposed algorithms.The experimental results show that: 1)Compared with existing community detection methods for signed networks,the CAEAw-SN algorithm is more adaptable and has lower time complexity,and the optimal solution is closer to the real solution.2)The CAEAq-SN algorithm not only retains all the advantages of CAEAw-SN,but also can further accelerate the convergence rate of Q-modurality,which make the algorithm find the optimal solution within less evolutions.