基于邻域关系的三支聚类方法  

Three-way Clustering Method Based on Neighborhood Relation

作　　者：花遇春 马建敏[1] HUA Yuchun;MA Jianmin(School of Science,Chang'an University,Xi'an 710064,China)

机构地区：[1]长安大学理学院,陕西西安710064

出　　处：《山西大学学报（自然科学版）》2023年第2期326-333,共8页Journal of Shanxi University(Natural Science Edition)

基　　金：国家自然科学基金(61772019;61603278)。

摘　　要：针对当前三支聚类方法不能有效处理数值型数据,且三支聚类结果受阈值影响问题,文章基于邻域关系提出了确定合适阈值的三支聚类方法。首先给出了确定最优K值的改进K-means聚类算法。进而基于邻域关系下的下、上近似引入精度,提出了权衡边界域和精度关系的有效性评价指标。应用该指标,给出了确定邻域下、上近似中最佳阈值的构建算法,进而得到三支聚类的核心域和边界域。最后,通过UCI数据集上的实验验证了该方法的可行性,且该方法有效提高了聚类精度。Since it is difficult for the existed thee-way clustering approaches to effectively deal with numerical data,and the results of three-way clustering are affected by the threshold,this paper proposes the three-way clustering with determined appropriate threshold based on the neighborhood relation.Firstly,an improved K-means clustering algorithm is proposed to acquire the optimistic K value.Then,the accuracy of lower and upper approximations is defined based on the neighborhood relation.The validity evaluation index is introduced to weigh the relationship between the boundary region and the accuracy.By using this index,an algorithm is constructed to determine the optimistic threshold in the lower and upper approximations based on the neighborhood.The corresponding core and boundary regions of the three-way clustering are obtained.Finally,an experiment on the UCI data is used to verify the validity of the proposed approach and the results of the experiment show that the approach can improve the accuracy of clustering.

关 键 词：三支聚类 K-MEANS聚类 邻域关系 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]