基于扩展Jarvis-Patrick聚类的异常检测算法优化及检测仿真  

Anomaly detection algorithm optimization and detection simulation based on extended Jarvis-Patrick clustering

作　　者：张利剑 陈晋鹏 ZHANG Lijian;CHEN Jinpeng(School of Electronics and Information,Xi’an Polytechnic University,Xi’an 710048,China)

机构地区：[1]西安工程大学电子信息学院,陕西西安710048

出　　处：《电子设计工程》2022年第13期100-104,共5页Electronic Design Engineering

基　　金：陕西省教育厅专项研究项目(19JK0361);西安市科技计划项目(GXYD7.16)。

摘　　要：针对异常检测聚类算法获得结果始终是零散的且小聚类数据量太大的问题,提出基于扩展Jarvis-Patrick聚类的异常检测算法优化。使用基于Jarvis-Patrick图的聚类方法进行检测。将数据抽象为点,计算两个点之间的距离并设置阈值以确定这两个点的相似性,通过两个点之间的相似性来确定两个点是否属于同一聚类。共享k最近邻关系点,获得扩展的共享k最近邻聚类子图以减少最终聚类结果中的聚类数量。使用提出的优化算法对KDD Cup99数据集进行实验,与传统Jarvis-Patrick聚类算法相比,提出算法有效提高了检测率,并且降低了数据量。Aiming at the problem that the results obtained by the anomaly detection clustering algorithm are always scattered and the amount of small clustering data is too large,an anomaly detection algorithm optimization based on extended Jarvis-Patrick clustering is proposed. The clustering method based on Jarvis-Patrick graph is used for detection. By abstracting the data as points,and then calculate the distance between the two points and set a threshold to determine the similarity of the two points.Determining whether the two points belong to the same cluster by determining the similarity between the two points. The k-nearest neighbor relationship points are shared,and an extended shared k-nearest neighbor cluster subgraph is obtained to reduce the number of clusters in the final clustering result. Use the proposed optimization algorithm to experiment on the KDD Cup99 data set. Compared with the traditional Jarvis-Patrick clustering algorithm,this algorithm effectively improves the detection rate and reduces the amount of data.

关 键 词：异常检测 Jarvis-Patrick聚类 扩展共享最近邻 归一化 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]