摘要
针对基于蚁群觅食原理的聚类算法初期收敛速度较慢的问题,以及未区分各维特征主次的缺陷,本文提出了一种两阶段蚁群聚类算法,以解决上述问题。第一阶段引入各只蚂蚁的实时信息素更新规则改善算法初期收敛速度较慢问题,并为第二阶段提供合理的初始隶属度矩阵;第二阶段利用隶属度矩阵自适应地赋予各维特征不同的权重,再用信息素强度和加权欧氏距离共同指导各只蚂蚁构造解。经过人工数据集和UCI数据集的测试,结果表明两阶段蚁群聚类算法可以加快算法初期收敛速度,同时提高聚类的准确率。
Focusing on the problem, which the clustering algorithm based on ant colony foraging principle of convergence may be slow in the initial stage, and the defects not distinguishing the various features of primary and secondary, this paper presents a two-stage ant colony clustering algorithm to solve the problems mentioned above. The first stage of algorithm which introduces the ant real-time initial pheromone update rule to improve the problem of low convergence speed in early algorithm, the second stage of algorithm, guiding the ants structural solution by the membership matrix to adaptively endow the reasonable feature weight of each dimension, as well as the pheromone intensity and weighted Euclidean distance . Through the test on artificial data set and UCI data sets, the results show that the t＇,~o-stage ant colony clustering algorithm can ~mprove the convergence speed in early algorithm, mean while, improve the accuracy of clustering.
出处
《洛阳理工学院学报（自然科学版）》
2017年第4期60-64,共5页
Journal of Luoyang Institute of Science and Technology：Natural Science Edition
关键词
蚁群聚类
两阶段
收敛速度
自适应权重
ant colony clustering
two stages
the rate of convergence
adaptive feature weight
