Abstract: To avoid local-optima and enhance the qualities of solutions, a hybrid evolutionary algorithms system was developed to perform data mining on databases represented as graphs. To increase the efficiency of the algorithm, a new substructure extension method based on single-label substructure extension was proposed, which could greatly reduce the times for performing graph isomorphism during the evolution. Experimental results on some typical data sets and theoretical proof indicate its high efficiency and correctness.

摘要： 将混合进化算法引入图数据挖掘,避免了陷入局部极值问题,提高了解的质量。在此基础上提出了一种基于单标签扩展的子结构扩展方法,该方法可以减少进化过程中图同构操作执行的次数。在典型数据集上的仿真实验和理论证明表明了该方法的高效性和正确性。

关键词: 进化算法, 图数据挖掘, 子结构发现, 子结构扩展, evolutionary computation, graphical data mining, substructure discovery, substructure extension