摘要： 现有入侵检测系统的效率和准确率较低。为此，提出一种基于主成分分析的特征提取方法。对数据源进行特征降维，将获得的主成分作为BP神经网络的输入数据进行识别。分析原始BP算法存在的问题，研究RPBP、CGBP、LMBP 3种改进BP算法，并进行仿真实验，结果表明，与原始BP算法相比，改进算法收敛速度快，漏报率和误报率低，能有效改善入侵检测的识别效果。
关键词:
入侵检测,
主成分分析,
神经网络,
BP算法,
误报率,
漏报率
Abstract: A feature extraction method using Principal Component Analysis(PCA) is proposed to improve the efficiency and accuracy of intrusion detection. This method reduces data dimensions and views some principal components as the inputs of BP neural network to finish data recognition. In order to overcome the problems in standard BP algorithm, three kinds of improved BP algorithm are studies and simulated, experimental results show that compared with standard BP algorithm, RPBP, CGBP and LMBP algorithm have well convergent speed and low false positive rate and false negative rate, they can improve recognition effect of three kinds of improved BP algorithm.
Key words:
intrusion detection,
Principal Component Analysis(PCA),
neural network,
BP algorithm,
false positive rate,
false negative rate
中图分类号:
李志清, 傅秀芬. 基于PCA的3种改进BP算法性能研究[J]. 计算机工程, 2011, 37(21): 108-110.
LI Zhi-Qing, FU Xiu-Fen. Performance Study on Three Kinds of Improved BP Algorithm Based on Principal Component Analysis[J]. Computer Engineering, 2011, 37(21): 108-110.