几种野值剔除准则在目标预测中的应用研究  

Research on Rules for Eliminating Outliers and Its Application to Target Prediction

作　　者：卢元磊[1] 何佳洲[1] 安瑾[1] 苗高洁[1] 

机构地区：[1]江苏自动化研究所,江苏连云港222006

出　　处：《指挥控制与仿真》2011年第4期98-102,共5页Command Control & Simulation

摘　　要：观测数据中的野值会影响目标预测的精度。分析比较了几种常用的野值剔除准则,包括最常用的3σ准则、奈尔准则、格拉布斯准则和狄克逊准则,将其应用到目标预测的数据预处理中,并比较了各种方法的野值剔除能力和对目标预测精度的影响。仿真结果表明:野值剔除准则的引入可有效减少数据中野值的数量,提高目标预测的精度。其中,格拉布斯准则的实用效果尤为明显,且误剔除率也得到了较好的控制。The outliers in the measured data will have a bad effect on the precision of target prediction.So how to eliminate outliers is an all-important problem.In this paper we first introduce several rules for eliminating outliers,and apply them to data preprocess in target prediction.By simulation on these rules,we demonstrate not only their capability of eliminating outliers but also how they impact the precision of target prediction.The results indicate that these rules can greatly improve the precision of target prediction,and the better the rule behaves,the more precise the prediction will be.This paper provides a good reference on how to choose appropriate rules for eliminating outliers while there is rare literature in this field up to now.

关 键 词：野值剔除 目标预测 准则 

分 类 号：TP274.2[自动化与计算机技术—检测技术与自动化装置]