基于灰色关联分析优化加权支持向量机的入侵检测系统的设计

Design of IDS Based on Based on Gray Relational Analysis and Optimization

针对传统SVM理论中现存的问题:没有考虑特征选取维数过高和特征重要性对分类结果影响的缺陷,提出了一种新的基于灰色关联分析优化的特征加权支持向量机算法。该方法首先利用灰色关联分析算法,筛选出主成分,再用对称不确定法计算各个特征对分类任务的权值。在充分的实验基础上将新方法应用于入侵检测中,得到了良好的实验结果。理论分析与数值实验都表明,尤其在小样本条件下,该入侵检测具有良好的检测率和更高的效率。

According to the existing problems in the theory of traditional SVM： the high dimensions of feature selection and the importance of characteristics in affecting the classification results are not considered. A new feature-weighted support vector machine algorithm is proposed, based on gray relational analysis and optimization. First, using the gray relational analysis algorithms to filter out the main ingredients, then using symmetrical uncertainty method to calculate the weights of the various characteristics of the classification task. In the full experiment on the basis of a new method used in intrusion detection, good experimental results is obtained. Theoretical analysis and numerical experiments indicate that our algorithm have good detection rate and higher efficiency in the intrusion detection, especially in the small sample experiment.