摘要
对于大规模入侵检测问题,分解算法是训练支持向量机的主要方法之一。在结构风险最小化的情况下,利用改进后的蚁群算法(QPSO)解决二次规划问题(QP),寻找最优解,并对ArraySVM算法进行了改进,同时对KDD入侵检测数据进行了检测。结果表明,算法精确度高于改进前的ArraySVM算法,并且减少了支持向量点数量。
One ofthe approaches to train support vector machine （SVM） for large.scale intrusion detection problem is the decomposition method. A new method based on the quantum-behaved particle swarm optimization （QPSO） is developed to solve quadratic programming （Qp） problem, and to find the optimal solution. ArraySVM algorithm is also improved to train KDD intrusion detection data sets. Based on the experimental results comparison and analysis, the present method is shown to more complete than previous ArraySVM algorithm in describing the modified algorithm precision, and also to reduce the number of support vector points.
出处
《计算机工程与设计》
CSCD
北大核心
2008年第1期34-36,共3页
Computer Engineering and Design
基金
国防应用基础研究基金项目(A1420061266)
教育部重点科学研究基金项目(105087)
