Training Samples Selection Method in Intrusion Detection System

Taking the example of designing classifier in intrusion detection system, this paper studies on samples selection problem for classifier and proposes a method fitting for large data set. First, use cluster analysis and the information known of classification to select boundary samples of each class. Then cluster for each class of the remaining non-border samples and adopt the method based on sample density to delete samples in each cluster. As reserving border samples and reducing training samples, it can guarantee generalization performance and training efficient of the classifier.