Research On Community Detection In Complex Networks Based On Intelligent Optimization Algorithm

There are many real-world complex systems with the characteristics of complex networks such as transportation systems, power grids and social relationship networks. Therefore, complex networks which can reveal the basic principle of complex systems have become the focus of current research. As a significant property of complex networks, community structure can find the hidden rules and functions. Because of this, it has received an enormous amount of attention. Community detection which is to find the community structure and then extract important information of each community structure has an important significant position in the complex networks field.In this thesis, community detection will be modeled as the objective optimization problem, based on which the single objective evolutionary algorithm and the multiobjective evolutionary algorithm are designed. In the single objective evolutionary algorithm, in order to overcome the drawbacks of single strategy and local optimal in the evolutionary process, community detection algorithm based on self-adapative strategy is proposed. In the algorithm the modularity is the objective function and crossover strategy pool and mutation strategy pool are designed, in which the strategies will be selected probabilistically based on statistical self-adaptive framework. Meanwhile the algorithm uses a hill-climbing strategy as the local search procedure that not only overcomes the drawback of local optimal but also improves the efficiency of the search. Experimental results show that the proposed algorithm achieves a better performance compared with similar approaches. In the multiobjective evolutionary algorithm, mutiobjective community detection algorithm based on non-dominated sorting is proposed in order to overcome the drawback that single objective evolutionary algorithm can only obtain one determined partition to the network. In the multiobjective algorithm two objective functions are put forward through which genetic operation and population updation procedure are set and local optimization function associationed with the two objective functions is defined for the local search. Experimental results show that the proposed algorithm can output a series of solutions through each run, this means more determined partitions to the network are given, which makes it possible to analyze the hierarchy structure of the network.