基于Voronoi和信息熵的空间离群样点检测  

Based on voronoi and information entropy spatial outliers detection algorithms

作　　者：王妍[1,2] 潘瑜春[1] 王慧 

机构地区：[1]国家农业信息化工程技术研究中心,北京100097 [2]首都师范大学信息工程学院,北京100037 [3]北京科络捷通讯产品有限责任公司,北京100043

出　　处：《计算机工程与设计》2010年第18期3998-4000,4008,共4页Computer Engineering and Design

基　　金：十一五科技支撑基金项目(2008BAJ08B03)

摘　　要：为了提高离群点挖掘的效率和准确度,在分析了传统离群点挖掘算法优、缺点的基础上,提出一种离群点检测算法。该算法利用Voronoi确定样点之间的邻近关系,通过参照邻域范围内其它样点的非空间属性值的信息熵作为离群因子,并根据离群因子标识出样点集中的离群点。以北京市大兴区土壤养分为例,实验结果表明,该检测算法能够高效、准确地检测出土壤样点中的离群点。In order to enhance the outlier mining efficiency and accuracy, soil outliers detection algorithm based on Voronoi and information entropy is proposed after analyzing the advantages and disadvantages of the classical outlier mining algorithms at first. The algorithm calculates information entropy of non-spatial attribute as the outlier factor by Voronoi neighborhood without parameters. Taking soil nutrient in Daxing district as an example, experimental results show that the proposed algorithm outperforms other existing algorithms in detection accuracy, user dependency and efficiency.

关 键 词：空间离群点 离群点检测 泰森多边形 信息熵 离散化 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]