摘要
群体智能优化算法Memetic算法(Memetic Algorithm,MA)采用进化算法的操作流程,引入局部搜索算子,使其在问题的求解中保证较高收敛性能的同时又能获得较高质量的解,克服了遗传算法等传统全局优化算法易"早熟"的问题,同时避免陷入局部解。在MA框架基础上,提出了全局动态适应MA算法,采用遗传算法为全局搜索算子,k-means算法为局部搜索算子。使用Java语言实现算法并对UCI中分类实验数据集进行测试,结果表明,将遗传算法和k-means结合的全局动态适应MA在分类问题中具有较高准确率。
Memetic Algorithm( MA) is one of the swarm intelligence optimization algorithms,and it has adopted the evolutionary algorithms operation process.The local search operator is used in MA to ensure the higher convergence in the solution of the problem and higher quality solutions can be obtained,then the problem that algorithm easy to"premature"of traditional global optimization algorithms,such as genetic algorithm,is overcame,at the same time,algorithm can avoid falling into local solution.The global dynamic adaptation MA is proposed based on the MA structure,the genetic algorithm is used as global search operator,k-means algorithm is used as local search operator.The algorithm is implemented by Java language and classification experiments data sets in UCI are tested,the results show that the global dynamic adaptation of MA,which combines genetic algorithm and k-means algorithm,has higher accuracy in the classification problem.
出处
《四川理工学院学报(自然科学版)》
CAS
2014年第5期43-46,共4页
Journal of Sichuan University of Science & Engineering(Natural Science Edition)