自适应的模糊谱聚类算法在文本聚类中的应用

Application of Auto-adaptation Fuzzy Spectral Clustering Algorithm in Text Clustering

传统聚类算法如k-means算法存在对样本空间形状敏感、一个样本点只能严格属于一个聚簇、需要人工指定聚簇数目等不足,这些不足之处都限制了文档聚类质量的提升。现有的模糊谱聚类算法只能解决前两个问题,而对于聚簇数目的自动确定却无能为力,因此本文提出一种自适应模糊谱聚类算法,该算法在模糊谱聚类的基础上引入自适应算法,解决聚类数目需要人工指定的问题。实验表明,将该方法用于文本聚类中可以取得较好的效果。

For traditional spectral clustering algorithms such as k-means algorithm, there exist a lot of deficiencies, for example its sensitivity on the shapes of the sample space, a sample point can only strictly belong to a cluster, need to specify the cluster number by manual work and so on, and these deficiencies limit the document clustering quality improvement. The existing fuzzy spectral clustering algorithm can only solve the first two problems, and the automatic determination of the number of clusters can not be determined. A kind of adaptive fuzzy spectral clustering algorithm was put forwards. The algorithm introduced the adaptive algorithm based on Fuzzy spectral clustering, which can solve the problem that the number of clusters should be specified manually. Experiments show that the proposed method can be used in text clustering and get the excellent effect.