自适应剪枝LS-SVM算法在入侵检测中的应用研究

摘要

针对SVM方法计算复杂度和时间复杂度较高的缺点,提出一种自适应剪枝LS-SVM算法.该算法通过块增量学习、剪枝过程以及逆学习的交替进行,大幅减少了支持向量的个数,降低了算法的计算复杂度和时间复杂度.实验结果表明,同标准C-SVM算法相比,应用该算法的入侵检测模型在检测时间、检测精度方面有着较好表现.%This paper proposes an algorithm of adaptive pruning LS-SVM against the shortcoming of high complexities in computation and time of the SVM. This algorithm reduces the number of support vectors through alternating the operation of chunk incremental learning,pruning and decremental learning,and reduces the computation and time complexity of the algorithm. The results of experiment show that the intrusion detection model using the proposed algorithm has quite good performances in detection time and accuracy compared with the standard C-SVM algorithm.