基于非对称相似度的文本聚类方法

文本聚类具有数据稀疏性的特点,常见的聚类方法采用基于距离的相异度,为了增强文档的区分特征,提出一种基于非对称相似度的方法,来度量文档对象之间的关联。定义了文本对象之间的非对称相似度度量。利用文本非对称相似度矩阵的稀疏特性,采用强连通构件的划分方法对文本对象进行聚类分析。并通过迭代的方法形成聚类结果的概念层次。实验结果表明:非对称相似度比距离相异度具有更高的准确率和更少的执行时间,当聚类结果簇数目达到较小时,准确率提高约为20%。

Text clustering based on asymmetric similarity

Text clustering data sets have sparse data spaces,with existing text clustering methods using distance-based dissimilarity to measure the document similarity.The document discrimination ability can be strengthened by a asymmetric similarity approach for text clustering.The asymmetric similarity is measured by a clustering analysis of the strong components of the sparse matrix.The approach provides a conceptual structure after the hierarchical clustering.Tests on textual data sets show that the asymmetric similarity measure provides higher precision with less run time than the distance-based dissimilarity method.With small numbers of clusters,the accuracy is improved by about 20%.