多层α-核散列聚类的异常数据社团发现算法

Abnormal Data Community Detection Algorithm of Multilayer α- Nuclear Hash Clustering

异常数据社团发现是解决异常数据挖掘中大规模未授权行为分析的重要方法,文中基于多层α-核心集对散列数据进行高斯核相似聚类,对大规模散列的网络入侵数据进行凝聚社团化处理,将难以聚类重叠的数据进行多层核簇类分析,并利用多层α-核心集处理与孤立点检测。通过实验证明,该方法对大数量具有明显散列特点的网络入侵攻击的检测具有较好的预期效果,在准确率与算法执行效率方面具有明显优势。

Abnormal data community detection is an essential method in solving the massive unauthorized behavior analysis in abnormal data mining. Based on multi-layerα-core set,the Gaussian kernel similar clustering of hash data is conducted,the community clustering done on the massive hash network invasion data and multi-layer core clustering analysis implemented on the data,which is hard to cluster together by using multi-layer core clustering,and also outlier detected by multi-layerα-core set. The experiment shows that this method has a good expected effect in of detecting network invasion attack with clear hash characteristics,and enjoys an obvious superiority in terms of accuracy and algorithm execution efficiency.