摘要
针对聚类划分问题,提出一种基于改进人工蜂群和最近邻原则的无监督聚类方法。该方法将每个蜜源作为聚类问题的一个可行解,设计了蜜蜂的多维编码结构。为了有效执行聚类,依据采蜜蜂和跟随蜂局部搜索阶段选择的较优聚类中心,利用k均值算法中的最近邻原则划分聚类空间的所有模式。为了使蜜蜂有较强的局部和全局搜索能力,根据聚类问题特点,提出了新的局部和全局搜索方法。仿真实验结果表明了新方法的可行性和高效性。
For the clustering partition issue, an unsupervised clustering approach based on improved artificial bee colony and nearest neighbour principle is given. This approach views every honey source as a candidate solution for the clustering, and designs a multidimensional code structure for the bee. To cluster effectively, based on better clustering centre selected by the employed bee and the onlookers in their local search phase, this approach divides all data patterns in clustering space by using the nearest neighbour principle in k-means. To improve local and global search ability of bee, the approach presents new local and global search method according to the feature of clustering problems. Simulative experimental results show that the new approach is feasible and effective.
出处
《计算机应用与软件》
CSCD
北大核心
2012年第12期65-68,共4页
Computer Applications and Software
基金
国家自然科学基金项目(61104179)
山东省高校智能信息处理与网络安全重点实验室(聊城大学)资助
聊城大学科研基金项目(X09034)
关键词
无监督聚类
人工蜂群
最近邻
K均值
粒子群优化
Unsupervised clustering Artificial bee colony Nearest neighbour k-means Particle swarm optimisation
作者简介
亓民勇，讲师，CCF会员（E200026622M），主研领域：智能优化算法应用，信息安全。
董金新，副教授。
