一种基于BIRCH的异常检测技术

An Anomaly Detection Method Based on Birch Algorithm

文章针对KNN存在的复杂度过高的问题,提出应用把BIRCH算法的层次聚类思想近似地计算weight的BirchOut算法,以降低其复杂度,同时利用孤立点挖掘的思想做异常检测.通过在KDD99数据集上的实验,我们验证了算法的有效性.

A new algorithm called BirchOut is proposed to reduce the complexity of KNM. calculation, which uses the idea of hiberarehy clustering,presents an anomaly detection method by using an outlier detection method. We apply this technique on KDD99 data set and get satisfactory results.