优化的RBF神经网络在入侵检测中的应用

在入侵检测的应用中,RBF神经网络训练样本的数据量比较大,但是训练中广泛应用的OLS方法存在大数据量训练时间过长、不能根据数据特性确定平滑参数的缺点。针对此问题该文采用了一种基于快速模糊C-均值算法(AFCM)和正交最小二乘法(OLS)算法相结合的AFORBF训练算法;试验证明,AFORBF算法解决了RBF在入侵检测系统中处理大数据量时间过长的问题,获得了较高的检测率,简化了网络结构,提高了网络性能。

Application of an optimized RBF neural network in intrusion detection

In intrusion detection applications,the training sample data based on the RBF neural network is quite big,but the OLS method widely applied in the training has some drawbacks for the training time of the large data amount is too long and the smoothing parameter can not be determined in accordance with data characters.This paper adopts the AFORBF training algorithm which is the combination of the fast fuzzy c-means algorithm(AFCM) and the orthogonal least squares(OLS) algorithm.Tests have proved that the AFORBF algorithm solves the problem of long time when the RBF deals with large data amount in the intrusion detection system,achieves a higher detection rate,simplifies the network structure and improves network performance.