基于相对密度的多耦合文本聚类算法

摘要

针对现有的空间向量模型在进行文本表示时忽略词条的位置和词条间关系的问题，提出了一种基于相对密度的多耦合文本聚类算法。在基于相对密度的聚类方法基础上，该算法根据相对密度越小文本相似性越小这一事实，将相对密度转换为文本相似度，融入了传统 DBSCAN 密度算法，并对核心对象的选取进行了优化。实验结果表明，与改进的 K-means 文本聚类和改进的 DBSCAN 文本聚类算法相比，本算法在文本聚类中更高效、聚类质量更优。%Aiming at the problem that existing space vector models ignore entry location and the relations between entries,this paper proposed a multi-coupled text clustering algorithm based on relative density.On the basis of clustering algorithms based on relative density and according to the reality that the smaller the relative density was,the smaller the text similarity was,the algorithm converted relative density to text similarity,integrated the traditional DBSCAN density algorithm,and optimized the selection of core objects.Experimental results show that,comparing with the improved K-means and the improved DBSCAN text clustering algorithms,the proposed algorithm performs more efficiently and has a better clustering quality in text cluste-ring.