Spark平台下综合属性权重离群点挖掘算法研究  

Research of Outlier Mining Algorithm Based on Comprehensive Attribute Weight on Spark Platform

作　　者：马晶[1] 刘建华 MA Jing;LIU Jianhua(Xi'an University of Posts and Telecommunications,Xi'an 710061;Information Center,Xi'an University of Posts and Telecommunications,Xi'an 710121)

机构地区：[1]西安邮电大学,西安710061 [2]西安邮电大学信息中心,西安710121

出　　处：《计算机与数字工程》2019年第8期1991-1995,共5页Computer & Digital Engineering

摘　　要：针对传统离群点检测算法对海量高维数据检测性能低下的问题,论文提出一种Spark平台下综合属性权重离群点挖掘算法S-CAWOM,通过对数据对象属性的综合加权处理,使不同的属性发挥不同的作用,并在Spark平台并行化提高算法效率。实验结果表明,该算法具有较高的性能。The traditional outlier detection algorithm is low efficiency in the high dimensional data detection performance.To solve this problem,a Outlier Mining Algorithm S-CAWOM is presented,which based on comprehensive attribute weight and to implement it under Spark platform.To make different attributes play different roles,the attributes of the data objects are weighted,and the algorithm on the Spark platform is parallelized to improve the efficiency of algorithm.Experimental results show that the algorithm has better performance.

关 键 词：离群点 综合属性权重 SPARK 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]