Internet intrusion detection system based on improved NN-SVM

Introducing the Degree of Class Ownership would improve NN-SVM algorithm in internet intrusion detection. According to the distance and the number of the same class or the different class, calculating the degree of class ownership of the sample point to its T nearest neighbors decided whether the sample point should be reserved or deleted. Based on this, the improved NN-SVM algorithm pruned the training sample set to reduce the confusion degree of the positive and negative categories. As a result, it could effectively reduce the cost of the learning and improve the generalization. The experiment shows that the improve NN-SVM algorithm, contrasting to traditional SVM algorithm, can effectively reduce the size of the training sample. So the improved NN-SV algorithm can solve the machine-learning problem with a small sample set and improve the performance of the system detecting.