顾及专题属性的空间聚类算法质量比较研究

Study on Quality Comparison of Spatial Clustering Algorithms Considering Thematic Attributes

顾及专题属性的空间聚类分析是空间数据挖掘中一个十分具有挑战性的研究课题。目前,顾及专题属性空间聚类分析一般可以分为划分的方法、层次的方法、基于模型的方法、基于密度的方法和混合的方法等。然而这些算法如何去比较质量、算法如何选取却是一个很少涉及却很重要的问题。本文对实际数据进行聚类分析实验,比较了11种顾及属性的空间聚类算法的质量,从整体质量差异和两两之间质量差异的角度分别进行了定性、定量分析,并分析了算法之间的相似度,这对于用户算法的选择、算法的评价具有指导意义。

Spatial clustering analysis,which takes account of thematic attributes,is a very challenging research topic in spatial data mining. At present,considering the thematic attribute,spatial clustering analysis can generally be divided into partition method,level method,model-based method,density based method and mixed method. However,how to compare the quality of these algorithms and how to select the algorithm is a very important problem. This paper makes a cluster analysis experiment on the actual data,and compares the quality of the 11 spatial clustering algorithms considering the attributes. The qualitative and quantitative analysis is carried out from the perspective of the overall quality difference and the quality difference between them,and the similarity between the algorithms is analyzed,which is of guiding significance for the user＇s choice of algorithm and algorithm evaluation.