CAJ | 학술논문

提出了一种构建轻量级的IP流分类器的wrapper型特征选择算法MRMHC-LSVM.该算法采用改进的随机变异爬山(MRMHC)搜索策略对特征子集空间进行随机搜索,然后利用提供的数据在无约束优化线性支持向量机(LSVM)上的分类错误率作为特征子集的评价标准来获取最优特征子集.在IP流数据集上进行了大量的实验,实验结果表明基于MRMHC-LSVM的流分类器在不影响分类准确度的情况下能够提高检测速度,与当前典型的流分类器NBK-FCBF相比,基于MRMHC-LSVM的IP流分类器具有更小的计算复杂度与更高的检测率.
제출료일충구건경량급적IP류분류기적wrapper형특정선택산법MRMHC-LSVM.해산법채용개진적수궤변이파산(MRMHC)수색책략대특정자집공간진행수궤수색,연후이용제공적수거재무약속우화선성지지향량궤(LSVM)상적분류착오솔작위특정자집적평개표준래획취최우특정자집.재IP류수거집상진행료대량적실험,실험결과표명기우MRMHC-LSVM적류분류기재불영향분류준학도적정황하능구제고검측속도,여당전전형적류분류기NBK-FCBF상비,기우MRMHC-LSVM적IP류분류기구유경소적계산복잡도여경고적검측솔.
This paper proposes a wrapper feature selection algorithm MRMHC-LSVM aiming at modeling lightweight flow classifiers by using the modified random mutation hill climbing (MRMHC) approach as the search strategy to specify candidate subsets for evaluation and using the linear support vector machine (LSVM) algorithm as the wrapper approach to obtain the optimum feature subset. The feasibility of the algorithm was examined by conducting several experiments on flow datasets. The experimental results show that a classifier based on the proposed approach can greatly improve the computational performance without negative impact on the classification accuracy. Further more, this approach is able not only to have smaller resource consumption, but also to have higher classification accuracy than the Nave Bayes method with Kernel density estimation after Fast Correlation-Based Filter (NBK-FCBF).