基于改进SVM的网络异常数据优化分类方法研究  

Network Anomaly Data Classification Method Research Based on Optimization of SVM

作　　者：吴浩[1] 王贤秋[2] 

机构地区：[1]内江职业技术学院,四川内江641100 [2]四川理工学院,四川自贡643000

出　　处：《科技通报》2016年第2期149-153,共5页Bulletin of Science and Technology

基　　金：四川理工学院项目编号(JG-1305)

摘　　要：对网络异常数据进行准确分类能够为网络入侵分类、保障网络安全提供准确的依据。传统算法没有考虑网络异常数据分布的不均衡性和高动态变化性,从而降低了分类的准确率和效率。为此,提出一种基于改进SVM的网络异常数据分类方法。在确定网络异常数据隶属度的时候考虑到其与类中心的关系,对传统的SVM进行了改进,在构建SVM分类器的过程中,引入了模糊隶属度函数,并将网络异常数据的分类问题转换为二次规划问题,最终实现网络异常数据的准确分类。仿真实验结果表明,利用改进算法进行网络异常数据分类,能够提高网络异常数据分类的准确率和分类效率,效果令人满意。The network abnormal data can be accurately classified as network intrusion classification, network security to provide accurate basis. Traditional algorithm does not consider the imbalance of the abnormal network data distribution and high dynamic change, which reduces the classification accuracy and efficiency. For this, put forward a kind of abnormal data classification method based on improved SVM network. Abnormal data in determining the network membership when considering its relationship with the center of the class, this paper improves the traditional SVM, in the process of constructing the SVM classifier, the introduction of the fuzzy membership functions, and network anomaly data classification problem into a quadratic programming problem, finally realizes the accurate classification of network anomaly data. Simulation experiment results show that the improved algorithm for network anomaly data classification, can improve the efficiency of network anomaly data classification accuracy and classification, the result is satisfactory.

关 键 词：改进SVM算法 网络异常数据 分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]