基于PCA的BP神经网络分类器_黄勤.pdfVIP

第23卷 第 7期 重 庆 工学 院学 报( 自然科学) 2009年 7月 Vol. 23 No. 7 Journal of Chongq ing Institute of Technology( Natural Science) Jul. 2009 PCA BP 1 1 2 1 黄 勤 , 常 伟 , 刘益良 , 刘衍鹏 ( 1. , 400030; 2. , 400050) : 由于入侵检测处理 多为高维数据, 为了提高入侵检测 效率和准确率, 提出了一种基于主成分分 析 (PCA) 特征提取方法, 对数据源进行特征降维, 将获得 主成分作为 BP神经网络 输入进行数据识别. 同 时介绍了 M atlab 中相关函数, 并与传统入侵检测方法进行了比较. 实验结果表明: 基于主成分分析 特征提取方 法在简化 BP神经网络规模 同时, 显著提高了入侵检测识别效果. : 入侵检测; 主成分分析;特征提取; BP神经网络; M atlab : TP393. 08 : A : 1671- 0924( 2009) 07- 0089- 08 BP N eural N etw ork C lassif ier B ased on Prin cipal Com ponent Analysis 1 1 2 1 HUANG Q in , CHANGW ei, LIU Y iliang, LIU Y anpeng ( 1. College ofAutom ation, Chongqing University, Chongqing 400030, Ch ina; 2. School of B iological Engineering, ChongqingUn ivers ity of Technology, Chongqing 400050, China) Abs tract: A feature extraction m ethod using principal component analysis is proposed to mi prove the efficiency and accuracy of intrusion detection, wh ich awl ays has high- dmi ensionaldata. Th ismethod reduces data dmi ensions and view s some principal components as the inputs of BP neural netw ork to fin ish data recogn ition. M eanw hile, this paper introduces som e correlative functionsused in MATLAB and makes a comparison w ith trad itional intrusion detection method. The expermi ental results show that the scale of BP neural network is smi plified and recognition effect is mi proved remarkably. K ey words: intrusion detection; feature extraction; principal component analysis; BP neural netw ork; M atlab