摘要
蚂蚁等群居式昆虫具有分布式、自组织、基于信息素间接通信(pheromone)等群体协作能力,模拟其智能行为的蚁群算法解决了许多复杂的问题并在并在数据聚类分析领域取得成效。本文首先介绍了基于蚂蚁的聚类算法的基本理论,讨论了参数σ对邻域平均相似度的影响并做了实验分析比较,然后提出利用离散点对算法进行改进,通过对离散点的检测算法能够对蚂蚁行为进行控制,使蚂蚁快速地决定下一个负载节点,从而有效地缩短聚类分折的执行时间。实验表明改进后的蚂蚁聚类算法具有较好的聚类特性,其收敛性也得到了有效改善。
Social insects such as ants have the ability of collaboration due to the swarm intelligence of ants and the mechanisms of their distributed behavior, self-organization and pheromone communication. Ant-based clustering has been applied in a variety of areas, such as problems arising in commerce, circuit design data clustering analysis in data-mining community. In this paper, we first present the basic theory of ant-clustering algorithm, discuss the scal- ing parameter which effect on neighborhood function, and analyze the experiment result. We also propose an im- proved algorithm based on outlier. The improved algorithm can check outlier to control the action of ants and decide the next load node quickly, then shorten the executive time and speed the convergence. At last, we compare our algo- rithm with related work and improve its effective.
出处
《计算机科学》
CSCD
北大核心
2005年第6期111-113,223,共4页
Computer Science
基金
重庆市自然科学基金(cstc.2004BB2086)
