一种高效的双边聚类集成算法

An Efficient Bilateral Clustering Ensemble Algorithm

谱聚类可以任意形状的数据进行聚类,在聚类集成中能够有效的提高基聚类的质量。以往的聚类集成算法中,聚类集成得到的结果并不是最终聚类结果,还需要利用聚类算法来获得最终聚类结果,在整个过程中会使得解由离散-连续-离散的转变。提出了一种基于谱聚类的双边聚类集成算法。算法首先在生成阶段使用谱聚类算法来获得基聚类,通过标准互信息来选取基聚类。将选出来基聚类和样本作为图的顶点,并对构建的图利用双边聚类算法对基聚类和样本同时聚类直接得到最终聚类结果。在实验中,将所提方法与一些聚类集成算法进行了比较,取得了较好的结果。

Spectral clustering can cluster data of any shape,which can effectively improve the quality of basic clustering in clustering ensemble.In previous clustering ensemble algorithms,the result obtained by clustering ensemble algorithm is not the final clustering result.It is also necessary to use the clustering algorithm to obtain the final clustering result.During the entire process,the solution will change from discrete to continuous to discrete.This paper proposes a bilateral clustering ensemble algorithm based on spectral clustering.Firstly,a spectral clustering algorithm was used to obtain base clusters in the generation stage,and the normalized mutual information was used to select the base clusters.The selected base clusters and samples were used as the vertices of the graph,and the bilateral clustering algorithm was used to cluster the base clusterings and samples simultaneously to obtain the final clustering results directly.In the experiments,the method in this paper was compared with some clustering ensemble algorithms,and the algorithm in this paper has achieved good results.