粗糙集属性约简的极限学习机网络入侵检测算法

ELM network intrusion detection algorithm based on rough set attribute reduction

针对网络入侵数据量大、属性冗余及属性之间线性相关导致分类算法计算速度慢、准确度不高等问题,提出一种改进粗糙集属性约简的极限学习机网络入侵分类算法。对训练集采用粗糙集正域和分辨矩阵相结合的方法获得属性核,筛选出只有属性核的数据集得到无冗余属性的特征集合;使用极限学习机(ELM)作为分类模型进行分类,使用支持向量机(SVM)、神经网络、极限学习机比较证明提出方法的有效性,为网络入侵检测提供一种新的解决方法。

Aiming at problem of slow computing speed and inaccuracy of the classification algorithm caused by large numbers of network intrusions,attribute redundancy and linear correlation between attributes,an extreme learning machine(ELM)network intrusion classification algorithm is proposed which is based on reduction of redundant attributes by the positive domain and discernibility matrixof the rough set.After the reduction of redundant attributes by the positive domain and discernibility matrix of the rough set,it gets the characteristic collection of non-redundant attributes.ELM serves as classification model.Through comparing it with SVM,neural network,ELM,the effectiveness of this method is proved,thus provide a new solution for network intrusion detection.