Community Detection Based On Evolutionary Muti-Objective Optimization And Implementation Of Prototype System

In recent years, as the emergece of complex network research, the Community Detecting in complex work have been given much attention. Community structrue is one of the most important topology attributes of complex network, which reveals its hidden rules and behavioral characteristic. Detecting community structure is crucial for uncovering the links between structures and function in complex networks. The popular methods are based on optimization of a single quality function, such as modularity. Many recent algorithms use the network modularity as quality metric, which turns the Community Detection into an optimization problem. However, for one thing, these algorithms have a high computational complexity, and thus they are not suitable for a complex network with a large size. For another, recent research reveals that these methods have the resolution limit.In order to solve the first problem, this paper first proposes a new Genetic Algorithm for Community Detection and then proves it a highly efficient algorithm via four experiments.To solve the second problem, this paper proposes to apply a multi-objective evolutionary algorithm to detect community. The method simultaneously optimizes two complementary objectives, and returns a set of different trade-off partitions. In order to find the best solution from the set of partitions, this paper further introduces four model selection criteria to select the best solution as candidate. The related experiments can be divided into two parts. The results of the first part illustrates that, in one run, the method accurately finds the communities at different hierarchical levels, which effectively avoids the resolution limit. In the second part, the experiments also show that the model selected by these two criteria is more accurate than GN and the algorithm based on modularity optimization.The last part of this paper is to design a prototype system for our algorithm. The main function of this system includes two parts: running the algorithm and visualizing the results.