摘要
在没有先验知识的前提下,聚类是分析样本集中不同类簇的有效方式。文中提出了一种基于改进力导向模型的聚类算法。为实现样本数据预处理的类内聚集和类间分离效果,设计了基于样本点局部密度和样本间距离的吸引力计算方法、基于样本点近邻连通图中边的介数的排斥力计算方法。实验结果表明,文中算法能够使得类内样本点更加聚集、类间样本点更加分离,可以有效地提高聚类的正确率。
Clustering is an effective way to analyze different clusters in sample sets without prior knowledge.In this paper,a clustering algorithm based on improved force-directed model is proposed.In order to achieve the effect of intra-class clustering and inter-class separation of sample data preprocessing,an attractive force calculation method based on local density of sample points and distance between samples is designed,and so is a repulsive force calculation method based on the betweenness of edges in the nearest neighbor connected graph of sample points.The experimental results show that the proposed algorithm can make the sample points of intra-class more clustering and the sample points of inter-class more separated,which can effectively improve the correct rate of clustering.
作者
刘风剑
刘向阳
LIU Feng-jian;LIU Xiang-yang(College of Science,Hohai University,Nanjing 211100,China)
出处
《信息技术》
2019年第10期51-54,58,共5页
Information Technology
基金
国家自然科学基金(61001139)
关键词
力导向模型
局部密度
聚类
边介数
force-directed model
local density
clustering
edge betweenness
