Journal of Mathematical Sciences: Advances and ApplicationsAuthor's: Shengkai Zhong
Pages: [1] - [20]
Received Date: March 22, 2016
Submitted by:
DOI: http://dx.doi.org/10.18642/jmsaa_7100121648
For the detection to modern networks intrusion, an intrusion detection model is put forward based on improved twin support vector machine. It employs new binary particle swarm algorithm to select parameters and feature subset. For the multi-classification problems, the least squares multi-classification twin support vector machine takes empirical risk minimization principle. And its generalization is optimized by adding rule item. The experiments use classic dataset called KDD’99. The result shows that the proposed algorithm is effective to solve the two categories and three categories problems of networks intrusion detection, improving the accuracy of classifiers.
feature selection, twin support vector machine, multi-class intrusion detection, particle swarm algorithm.
