摘要
异常流量的准确识别在网络安全中起着重要作用,支持向量机(Support Vector Machine,SVM)已经成功地应用于分类和函数逼近等方面,而核函数参数和惩罚参数(C)的选取对SVM的分类性能起着关键作用.为了提高SVM的分类性能,提出一种基于改进蚱蜢算法优化SVM的异常流量识别方法,命名为SAGOA-SVM.在对蚱蜢算法进行实验研究后发现其局部搜索能力较弱,本文通过引入模拟退火算法和位置偏移机制增强蚱蜢趋向食物源的随机性来改进蚱蜢算法优化SVM参数的性能,从而提高SAGOA-SVM算法对异常流量的识别率.在选取的7个标准UCI数据集上的实验结果表明,所提出的SAGOA-SVM算法有很好的分类精度和性能.
Accurate identification of abnormal traffic played an important role in network security. Support vector machine( SVM) was successfully applied to classification and function approximation. However,the selection of kernel function parameters and penalty parameters( C) played a key role in SVM classification performance. To improve the classification performance of SVM,an abnormal traffic identification method based on improved grasshopper optimization algorithm for optimizing SVM is proposed,named SAGOA-SVM. An experimental study of grasshopper algorithm show that its local search ability was weak. In order to enhance the local search ability of the algorithm,the simulated annealing algorithm( SA) and the position offset mechanism were introduced.Therefore,the performance of the algorithm for searching SVM parameters was improved. Experimental results on selected 7 standard UCI datasets show that the proposed SAGOA-SVM algorithm has good classification accuracy and performance.
作者
吕赵明
张颖江
Lv Zhaoming;Zhang Yingjiang(School of Computer Science,Hubei University of Technology,Wuhan 430068,China)
出处
《湖南科技大学学报（自然科学版）》
CAS
北大核心
2019年第4期90-96,共7页
Journal of Hunan University of Science And Technology：Natural Science Edition
基金
教育部下一代互联网创新项目基金资助(NGII20150404)
关键词
蚱蜢优化算法
模拟退火算法
支持向量机
核函数
异常流量识别
grasshopper optimization algorithm
simulated annealing algorithm
support vector machine
kernel function
abnormal traffic identification
