高维数据集之中基于距离的离群快速检测算法  

Fast Mining of Distance-based Outliers in High-dimensional Datasets

作　　者：乔天成 QIAO Tian-cheng(Taiyuan Station of the First Communication Station, PLA Army General Staff, Taiyuan 030012 China)

机构地区：[1]中国人民解放军陆军参谋部第一通信站太原机务站,山西太原030012

出　　处：《科技创新与生产力》2017年第11期67-71,共5页Sci-tech Innovation and Productivity

摘　　要：针对现有的挖掘算法并不适用于大规模的高维数据集的问题,给出了一种针对高维数据集的RBRP算法,能够快速检测出数据集中基于距离的异常,该算法将对数线性作为数据点个数的函数,线性作为维数的函数。实验结果表明,RBRP算法始终优于ORCA算法,且是一种针对高维数据集的最优的基于距离的异常检测算法,并且RBRP算法的优势往往超过ORCA算法一个数量级。The existing algorithms for mining distance-based outliers are not suitable for large and high-dimensional data sets.Based on the problem,this paper presents RBRP,a fast algorithm for mining distance-based outliers.T he algorithm takes log-linear as a function of the number of data points,and linear as a function of dimension.Experimental results show that it is better than ORCA algorithm and its advantages often exceed the ORCA algorithm by an order of magnitude.

关 键 词：数据挖掘 算法 离群 高维数据集 近似K-近邻 聚类 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] TP301.6[自动化与计算机技术—计算机科学与技术]