支持向量機(SVM )作為一種分類技術(shù)已經(jīng)成功運用于入侵檢測，但是支持向量機的
性能與參數(shù)的選擇相關(guān)。在實際應(yīng)用中支持向量機的參數(shù)選取問題一直沒有得到很好地解決。粒子群優(yōu)化（PSO）算法作為一種基于群智能方法的進化計算技術(shù)，具有良好的全局搜索能力。為了能夠自動獲取最優(yōu)的支持向量機參數(shù)，提出了在入侵檢測系統(tǒng)中基于改進微粒群優(yōu)化(IPSO)算法的支持向量機參數(shù)選擇方法，以kdd99數(shù)據(jù)集進行了仿真實驗。仿真實驗結(jié)果表明：基于粒子群訓(xùn)練的支持向量機方法能夠比較好地提高入侵檢測系統(tǒng)中數(shù)據(jù)的分類精度。
關(guān)鍵詞：粒子群優(yōu)化；支持向量機；入侵檢測系統(tǒng)
Abstract : As a classification technical , support vector machines（SVM）have been applied in intrusion detection successful, But the performance of SVM is determined by its hyper parameters. In practice, the problem on how to select parameters of SVM is not solved properly. As an evolutionary computation technique based on swarm intelligence particle swarm optimization (PSO) algorithm has high global search ability, In order to optimize parameters of SVM automatically, a parameter selection approach based on PSO is proposed in this paper. Experiments with the data set kdd99 show that the method which based on PSO training of the SVM can improve the classification accuracy of dataset in IDS.
Key words: Particle swarm optimization; support vector machines; Intrusion detection system