基于文本最小相似度的中心选取方法  

An Approach to Center Selection Based on Minimal Similarity Among Texts

作　　者：郑伟[1] 季铎[1] 蔡东风[1] 王爽[2] 

机构地区：[1]沈阳航空工业学院自然语言处理研究室,辽宁沈阳110034 [2]沈阳航空工业学院北方软件学院,辽宁沈阳110034

出　　处：《广西师范大学学报（自然科学版）》2008年第3期198-201,共4页Journal of Guangxi Normal University:Natural Science Edition

基　　金：国家863计划基金资助项目(2006AA01Z148);教育部科学技术研究重点项目(207148)

摘　　要：基于划分的聚类算法是一种局部最优算法。聚类初始中心的选择对该聚类算法的收敛速度和聚类的性能都有很大的影响。初始中心点应该选择来自不同的类并且初始中心点文本之间的相似度应尽量小,为此提出了一种新的基于最小相似度的中心选取方法,该方法选择相似度最小的两个样本分别作为初始的两个中心,然后依次选择到已知中心相似度最小的样本作为其他类的中心。实验表明该方法能够选出来自不同类的样本作为聚类的初始中心,同其他初始中心选择方法比较,聚类算法的性能有明显提高。Clustering algorithms based on partition is a local optimization algorithm. The .selection of starting center points of clustering has great effects on the constringency speed of this clustering algorithms and the performance of clustering. The starting center points should select from different classes and the similarity of starting center points text should be as small as possible. So,this paper proposes a new method based on the center of sum function of the minimal similarity. The method selects two samples which have the minimal similarity as two centers of initial clustering,then selects the sample which has the minimal similarity to the selected points as other classes＇ center. Experiments show this method can select the samples from different classes as the starting point of clustering, compared with other methods ,the performance of clustering algorithms is obviously improved.

关 键 词：K—Means 最小相似度 文本聚类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]