考虑空间格局的谱聚类算法及其应用  

Spectral Clustering Algorithm Considering Spatial Pattern and Its Application

作　　者：于永玲[1] 李向[1] 宗思生[1] 施进发[1] 

机构地区：[1]郑州航空工业管理学院计算机科学与应用系,河南郑州450015

出　　处：《河南科技大学学报（自然科学版）》2013年第5期101-104,10,共4页Journal of Henan University of Science And Technology:Natural Science

基　　金：河南省科技厅科技攻关基金项目(122102110179;112102210352;122102210200);河南省教育厅自然科学研究基金项目(2011A630043;2012B520062);河南省软科学基金项目(132400410621)

摘　　要：采用基于划分的空间聚类方法对地理要素进行聚类时,若仅考虑属性数据,得到与实际空间分布不相符的聚类结果。提出一种考虑空间对象属性特征和空间位置关系的谱聚类方法,首先,计算空间对象的局部离群指数,结合空间格局将样本中的异常点剔除,然后以空间临近为约束条件进行谱聚类分析。以包头地区土壤重金属形态数据为例进行聚类分析,分析结果表明:该方法能够克服谱聚类对初始聚类中心敏感的问题,既能反映属性特征数据的相似程度,又能反映对象的空间分布特性,对空间对象的聚类分析效果优于传统的谱聚类算法。There will be some differences between clustering results and the actual situation if we only consider attribute data of spatial objects based only on partition with geographic features. In this paper,a spectral clustering algorithm including spatial pattern was presented for spatial clustering analysis. To deal with the abnormal points in the data,the local outlier factors were calculated firstly. The spatial neighbor was used as constrains of spatial clustering. And the the abnormal points were removed and judged by spatial pattern. To verify the effectiveness of the algorithm,the improved algorithm was applied to solving environmental quality assessment of soil heavy metal form data with the city of Baotou taken as an example. The results show that the method can not only overcome sensitity of the spectrum clustering in the problem of initial clustering center but also reflect the degree of similarity of the data attributes and the characteristics of spatial distribution. The result of spatial clustering method is better than that of traditional spectrum clustering algorithm.

关 键 词：空间格局 谱聚类 邻接矩阵 重金属形态 土壤 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TP399[自动化与计算机技术—控制科学与工程]