微型电脑应用

针对网络入侵检测中数据集的不均衡性和异构性,提出了一种基于密度均衡的网络入侵检测算法,利用异构距离计算类密度,依据类密度之间的关系把多数类数据集划分为多个子集;然后依据每个子集密度与少数类密度之间的关系动态计算重取样后样本数目,利用K-means聚类算法对每个子类进行重取样,实现数据集的均衡化处理。最后在KDDCUP99数据集上进行仿真实验,实验结果表明,该算法提高支持向量机对少数类的分类性能,同时提高运行效率。

In order to reduce the impact of imbalanced and Heterogeneous data for dataset of network intrusion detection,an intrusion detection algorithm based on density equalization is proposed.The algorithm calculates the class density according to Heterogeneous distance.Then majority class is divided into multiple sub classes according to the relation between sample density of particle density and minority class.Calculate the value of K and resampling for every sub classes with K-means cluster algorithm.Then get the balance data sets.Finally,experiment results with KDDCUP99 dataset show the algorithm can improve the classification performance at imbalanced dataset,especially for the minority class samples.