基于Kohonen神经网络算法的网络入侵聚类算法的测试研究

摘要

This paper presents a method of clustering of network intrusion based on Kohonen network.Firstly,the basic theory,principle and algorithm steps are introduced.Then,matlab software platform was used for testing the proposed network intrusion algorithm.Finally,this algorithm was compared with other methods though simulation tests.Experimental results show the Kohonen neural network clustering algorithm is better than PNN algorithm in three aspects,i.e.,training accuracy,testing accuracy and operation time,its accuracy rate can reach 98.1％.%提出一种基于Kohonen网络的网络入侵聚类研究的方法,在阐述基本理论、原理和算法步骤基础上,利用Matlab软件平台对提出的网络入侵算法进行测试研究,并同其他方法进行仿真对比,发现Kohonen神经网络算法的网络入侵聚类在训练准确率、测试准确率和运行时间3个方面都优于PNN算法,其准确率可以达到98.1％.