入侵检测系统中基于凝聚聚类算法的训练样本集的构造

Construction of the Training Data Set in Intrusion Detection Systems Based on the Agglomerate Clustering Algorithm

本文提出了一种入侵检测系统中训练样本集的构造方法,首先通过保留边界样本和删除内部样本进行样本选择,然后使用遗传算法与凝聚聚类算法相结合的方法对样本数量较少的类构造虚拟样本。这样得到的训练子集样本数量少,而且分布均匀。

A constructing method is proposed to train the data set in intrusion detection systems in this paper. First, through retraining the boundary and discarding most of the interior training samples, we select important samples from classes containing more samples,construct virtual samples for the class that contains less samples in the way of combining the genetic algorithm and the agglomerate clustering algorithm. The number of the training subset samples gained in this way is few and its distribution is symmetrical.