利用改进遗传算法进行复杂网络社团发现

摘要

针对使用遗传算法进行复杂网络社团发现时，存在较强随机性以及容易陷入局部最优解的缺陷，提出一种基于遗传框架的复杂网络社团发现新方法。其通过一次迭代标签传播方法进行种群初始化，针对字符串表示法交叉困难的特点提出了统一标签交叉策略，并采用有指向性的变异策略解决遗传算法随机变异的缺陷问题。实验结果表明：对典型的人工生成网络结构和真实网络结构，该方法能够较准确地发现社团结构；与经典算法进行比较，该方法具有较高的社团发现精度且收敛速度较快。%Against the defects of stronger randomness and local optimal solution when the community detection in complex networks was made using the genetic algorithm, a new method of community detection in complex networks was presented using the genetic algorithm structure. The single-iteration label propagation method was utilized to make population initializing, a strategy of unified label crossover against crossing difficulty of string representation was proposed, and the directional mutation strategy was adopted to solve the defects of random mutation in the ge-netic algorithm. The experimental results show that for the typical computer-generated network structure and real-world network structure, the method can detect community structure more accurately. Compared with the classical algorithms, it has a higher precision of community detection and a faster convergence speed.