基于单元格和属性权重的离群点检测  

OUTLIER DETECTION BASED ON CELL UNIT AND ATTRIBUTE WEIGHTS

作　　者：姜立明[1] 柴瑞敏[1] 

机构地区：[1]辽宁工程技术大学电子与信息工程学院,辽宁葫芦岛125105

出　　处：《计算机应用与软件》2012年第10期216-218,244,共4页Computer Applications and Software

摘　　要：离群点检测是数据挖掘领域的一个重要的研究方向。针对高维数据空间中离群数据的挖掘速度和准确度的问题,提出一种基于单元格的离群点检测算法。该算法在高维数据空间中对数据进行降维,并且将数据依据属性权重划分成若干空间单元,从而减少查询次数,提高离群数据的挖掘速度。另外,通过对属性的加权处理能够更有效地突出属性的特殊性,从而提高挖掘的准确度。理论分析和实验结果表明了该方法是有效可行的。Outlier detection is an important research direction in data mining field. A cellbased outlier detection algorithm is proposed ac cording to the problems of mining speed and accuracy of outlier data in high dimension. In the algorithm the data is conducted dimensionality reduction in high dimensional data space and is divided into a number of spatial units based on attribute weights, thereby reduces the number of queries and increases the mining speed on outlier data. In addition, the weighted processing on properties can more effectively highlight the specificity of attributes, thus improves the accuracy of mining, Theoretical analysis and experimental results all show that the method is feasi ble and effective.

关 键 词：数据挖掘 离群数据 单元格 属性权重 粗糙集 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]