聚类融合算法的实验评价方法

The Evaluation Method of Clustering Fusion Algorithm

聚类融合算法被认为是数据分析的有效工具之一。然而,除了研究优良的聚类融合算法,如何评价聚类的质量也被认为是难题之一。传统的外在方法使用专家评价的基准作为参照。但是实际上,这种基准不但昂贵,而且常常不容易得到。因此,一种新颖的基于实验的聚类融合算法评价方法被提出,其参照基准是基于所有聚类融合算法折衷所得出来的。基于这个方法的设计框架,实验部分使用了SLC(single-linkage clustering)和IVC(iterative voting clustering)在2个仿真和3个UCI数据集上进了评价对比,并将结果和传统外在方法进了比较。从传统外在方法看来,当参与评价的算法是强聚类融合算法时,该评价方法结果与传统方法的评价结果一致。

Clustering fusion algorithm is considered to be one of the effective tool of data analysis. However, in addition to excellent Clustering Fusion Algorithm Research, how to evaluate the quality of clustering is also considered one of the problems. Expert evaluation using external traditional methods as a reference benchmark. But in fact, this benchmark is not only expensive, and often not easy to get. Therefore, an experimental clustering fusion algorithm is proposed based on a novel evaluation method, the reference is all Clustering Fusion Algorithm Based on the compromise. This method is based on the design framework, experiments using SLC （single-linkage clustering） and IVC （iterative voting clustering） in 2 Simulation and 3 UCI data sets up a comparative evaluation, and the results were compared with traditional external method. From the point of view of the traditional external method, when participating in the evaluation algorithm is clustering fusion algorithm, the evaluation results of the evaluation method and the traditional method of the same.