A novel SVM by combining kernel principal component analysis and improved chaotic particle swarm optimization for intrusion detection

A novel support vector machine (SVM) model by combining kernel principal component analysis (KPCA) with improved chaotic particle swarm optimization (ICPSO) is proposed to deal with intrusion detection. The proposed method, in which multi-layer SVM classifier is employed to estimate whether the action is an attack, KPCA is applied as a preprocessor of SVM to reduce the dimension of feature vectors and shorten training time. To shorten the training time and improve the performance of SVM, N-RBF is employed to reduce the noise generated by feature differences, and ICPSO is presented to optimize the punishment factor C, kernel parameters $$\sigma $$ ? and the tube size $$\varepsilon $$ ? of SVM, which introduces chaos optimization and premature processing mechanism. Experimental results illustrate that the improved SVM model has faster computational time and higher predictive accuracy, and it can also shorten the training time and improve the performance of SVM.