Network intrusion classification based on extreme learning machine

Extreme learning machine (ELM) is an efficient learning algorithm which can be easily used with least human intervene. But when ELM is applied as multiclass classifier, the results of some classes are not satisfactory and it´s hard to adjust the parameters for these classes without affecting other classes. To overcome these limitations, a novel method is proposed. In proposed approach, binary ELM classifiers for each class are combined into an ensemble classifier using one-to-all strategy. The experiment on NSL-KDD data shows that the proposed approach outperforms ELM multiclass classifier, decision tree, neural network (NN) and support vector machines (SVM).