基于地统计学的空间离群点检测算法的研究  

Geographical statistics based approach for spatial outlier detection

作　　者：刘莘[1,2] 张绍良[2] 王飞[3] 张赛男[4] Liu Xin Zhang Shaoliang Wang Fei Zhang Sainan(School of Medicine Information, Xuzhou Medical College, Xuzhou Jiangsu 221004, China School of Environment Science ＆ Spatial Informatics, China University of Mining ＆ Technology, Xuzhou Jiangsu 221008, China China Telecom Co. Xuzhou Branch, Xuzhou Jiangsu 221000, China College of Command Information Systems, PLA University of Science ＆ Technology, Nanfing 210007, China)

机构地区：[1]徐州医科大学医学信息学院,江苏徐州221004 [2]中国矿业大学环境与测绘学院,江苏徐州221008 [3]中国电信股份有限公司徐州分公司,江苏徐州221000 [4]中国人民解放军理工大学指挥信息系统学院,南京210007

出　　处：《计算机应用研究》2016年第12期3700-3704,共5页Application Research of Computers

基　　金：国家"十二五"科技支撑计划资助项目(2011BAC08B03);江苏省煤基CO2捕集与地质储存重点实验室(中国矿业大学)开放基金资助项目(2015A01);江苏高校优势学科建设工程资助项目(SZBF2011-6-B35)

摘　　要：针对传统空间离群点检测算法构建邻域时参数选择困难、处理高维数据的时间复杂度较高等问题,提出了一种基于地统计学的空间离群点检测算法。该算法将空间自相关理论引入空间离群检测中,首先利用3σ规则识别全局离群点,然后利用Delaunay三角网构建空间邻域,将邻域节点均值代替全局离群点,最后使用局部Moran’I作为空间异常的度量方法。仿真结果表明,该方法不需要选择参数,鲁棒性较强、检测率较高、误警率较低。In view of the traditional spatial outliers detection algorithms have difficuh in parameter selection when constructing neighborhood and have high time complexity when dealing high-dimensional data,this paper proposed a spatial outliers detection algorithm based on the geographical statistical theory. This algorithm introduced the spatial autocorrelation theory into the spatial outlier detection, firstly used 3σrules to identify global outliers, and then used Delaunay triangulation neighborhood relationship building spatial neighborhood relationship, instead the global outliers by the average neighborhood node, finally used local Moran＇ I as space abnormal measure method. The simulation results show that the method does not have to select parameters, strong robustness, higher detection rate and lower false alarm rate.

关 键 词：地统计学 空间离群点 DELAUNAY三角网 局部Moran’I 

分 类 号：TP301.5[自动化与计算机技术—计算机系统结构]