基于属性相似性的超图聚类改进算法

摘要

Clustering can be applied to many aspects of modern life,while data objects in modern life are often high dimensional and sparse. For this kind of high-dimensional data,the traditional clustering algorithm cannot be effectively dealt with. This paper proposes a improved algorithm of hypergraph clustering based on attribute similarity, on the basis of the original algorithm of hypergraph clustering, According to the threshold-edge distance form the hypergraph model and the hypergraph partitioning method to cluster the data object,using cluster-heads singular eigenvalue to evaluate the quality of clustering.%聚类可应用于现代生活的诸多方面，现代生活中的数据对象往往是高维、稀疏的。对于此类高维数据，传统聚类算法不能有效地处理。提出一种基于属性相似性的改进的超图聚类算法，在原有超图聚类算法的基础上，根据超边距离阈值形成超图模型并采用超图分割法对数据对象进行聚类，采用簇内奇异特征值进行评估聚类质量。