最大距离法选取初始簇中心的 K-means 文本聚类算法的研究
由于初始簇中心的随机选择,K—means算法在聚类时容易出现聚类结果局部最优、聚类结果不稳定、总迭代次数较多等问题。为了解决K.means算法所存在的以上问题,提出了最大距离法选取初始簇中心的K.iTleans文本聚类算法。该算法基于这样的事实:距离最远的样本点最不可能分到同一个簇中。为使该算法能应用于文本聚类,构造了一种将文本相似度转换为文本距离的方法,同时也重新构造了迭代中的簇中心计算公式和测度函数。在实例验证中,对分属于五个类别的1500篇文本组成的文本集进行了文本聚类分析,其结果表明,与原始的K—means聚类算法以及其他的两种改进的K—means聚类算法相比,新提出的文本聚类算法在降...
Saved in:
Published in | 计算机应用研究 Vol. 31; no. 3; pp. 713 - 715 |
---|---|
Main Author | |
Format | Journal Article |
Language | Chinese |
Published |
School of Information Science & Technology,Southwest Jiaotong University,Chengdu 610031,China
2014
Engineering School,Tibet University,Lhasa 850000,China%School of Information Science & Technology,Southwest Jiaotong University,Chengdu 610031,China%Engineering School,Tibet University,Lhasa 850000,China |
Subjects | |
Online Access | Get full text |
ISSN | 1001-3695 |
DOI | 10.3969/j.issn.1001-3695.2014.03.017 |
Cover
Summary: | 由于初始簇中心的随机选择,K—means算法在聚类时容易出现聚类结果局部最优、聚类结果不稳定、总迭代次数较多等问题。为了解决K.means算法所存在的以上问题,提出了最大距离法选取初始簇中心的K.iTleans文本聚类算法。该算法基于这样的事实:距离最远的样本点最不可能分到同一个簇中。为使该算法能应用于文本聚类,构造了一种将文本相似度转换为文本距离的方法,同时也重新构造了迭代中的簇中心计算公式和测度函数。在实例验证中,对分属于五个类别的1500篇文本组成的文本集进行了文本聚类分析,其结果表明,与原始的K—means聚类算法以及其他的两种改进的K—means聚类算法相比,新提出的文本聚类算法在降低了聚类总耗时的同时,F度量值也有了明显提高。 |
---|---|
Bibliography: | 51-1196/TP ZHAI Dong-hai, YU Jiangi, GAO Fei2, YU Leit, DING Feng2 ( 1. School of Information Science & Technology, Southwest Jiaotong University, Chengdu 610031, China; 2. Engineering School, Tibet Univer- sity, Lhasa 850000, China) K-means clustering algorithm; maximum distance ; text clustering; text distance ; measurement function ; F-measure Due to the random selection of initial cluster centers, K-means clustering algorithm is prone to local optimal and in- stability of clustering results, and huge number of iterations. To overcome the above problems, this paper selected the initial cluster centers according to maximum distance, and it was based on the fact that the farthest samples were the least likely in the same cluster. To apply the improved algorithm into text clustering, it constructed a method to transform text similarity into text distance, and also reconstructed cluster center iteration formula and measurement function. It employed a text set which included 5 categories and 1 500 texts in the exp |
ISSN: | 1001-3695 |
DOI: | 10.3969/j.issn.1001-3695.2014.03.017 |