基于自组织映射的遗传聚类算法  

Hyper-heuristic Genetic Cluster Algorithm Based on Self-organizing Map

作　　者：徐聪[1] 黄文准[1] 黄世奇[1] 

机构地区：[1]西京学院信息工程学院,陕西西安710123

出　　处：《计算机与现代化》2017年第4期38-43,共6页Computer and Modernization

基　　金：陕西省教育厅科研项目(16JK2238);西京学院科研基金资助项目(XJ150119)

摘　　要：遗传聚类算法往往需要较大的种群规模才能得到最优解,导致收敛速度慢,针对这一问题,本文提出一种基于自组织映射的超启发遗传聚类算法。首先利用自组织映射把数据空间转换到特征空间,再在特征空间里利用遗传算法进行搜索,然后进行反映射,即把聚类结果在数据空间里表现,从而得到一组解,同时利用K-means算法在数据空间里进行粗聚类,获得另一组解,最后比较2组解的聚类结果,相同的样本保留,不同的再次聚类,进而有效地保证了最优解的获得。计算机仿真实验验证了所提算法在种群规模较小的情况下,可以获得较高的准确率。Genetic clustering algorithm can obtain the optimal solution on the condition of the larger population,however,which leads to a slow convergence speed. In order to tackle the challenge problem,this paper proposes a novel hyper-heuristic genetic cluster algorithm based on self-organizing map. Firstly,the data space is converted to feature space by exploiting self-organizing map method. Secondly,one solution can be found by employing genetic algorithm in the feature space to diminish the computational load of the presented algorithm. Then,the one solution can be reflected to the data space. Moreover,another solution can be found by the K-means algorithm in the data space. The optimal solution is obtained according to the optimal method,which the same cluster results maintained and the different ones are clustered again to further ensure optimal solution. The extensive simulation results demonstrate the proposed algorithm has much higher accurate rate in the case of small population in comparison with genetic clustering algorithm.

关 键 词：遗传算法 自组织映射 聚类分析 超启发搜索 图像分割 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]