论文部分内容阅读
针对当前主流的图形数据挖掘算法常采用的贪婪式查找带来的易陷入局部最优这一问题,将进化算法引入图形数据挖掘.以增强算法的全局查找能力.定义基于图形数据的交叉和变异算子.考虑到进化算法局部搜索能力弱的特点,在变异算子的设计中融入爬山算法的思想,以进一步提高解的质量.另外还改进原算法针对某一特定子结构的实例收集方法.实验表明,以上措施增强系统对假设空间的查找能力,提高解的质量.
Aiming at the problem that the greedy lookups often lead to the local optimum caused by the current mainstream graph data mining algorithms, the evolutionary algorithm is introduced into the graph data mining to enhance the global search ability of the algorithm.Definition of the crossover and mutation based on the graph data Considering the weakness of local search ability of evolutionary algorithm, the idea of hill-climbing algorithm is incorporated into the design of mutation operator to further improve the quality of the solution.In addition, the original algorithm is also improved for instance collection method for a particular substructure. Experiments show that the above measures enhance the ability of the system to find the hypothesis space and improve the quality of solution.