基于混沌免疫聚类的异常检测算法  

Anomaly Detection Algorithm Based on Chaos Immune Clustering

作　　者：翁鹤[1] 王喆[2] 

机构地区：[1]中国空空导弹研究院信息部 [2]洛阳电光设备研究所瞄准显示系统研发部,河南洛阳471000

出　　处：《软件导刊》2016年第4期55-58,共4页Software Guide

摘　　要：在对免疫聚类类型和流程分析的基础上,结合混沌变量的遍历性和聚类融合方法,提出了基于改进aiNet(artificial immune net)聚类算法CO-aiNet(Chaos optimization of artificial immune net)的异常检测算法ICDA(Immune clustering based Anomaly detection)。CO-aiNet算法在引入权重矢量、相关度等基础上,采用模拟退火算法和概率准则优化监测数据,优化了聚类效果;引入异常因子概念,通过对多次聚类融合后的数据按照异常标记次数排序,得到异常数据集。实验表明,CO-aiNet算法聚类效果优于同类算法,基于聚类融合的异常检测准确性和稳定性显著提升。Based on the type of immune clustering process and analysis, joint with ergodic chaotic variables and clustering fusion, proposedan enhanced aiNet （artificial immune net） clustering algorithm CO-aiNet （Chaos optimization of artificial immune net） applied on an anomaly detection algorithm ICDA（Immune clustering based Anomaly detection） is recommen- ded. CO-aiNet algorithm introduces the concept of weight vectors, relevancy, meanwhile, by using the simulated annea- ling algorithm and probability criterionto optimize the monitored data, it has further optimized the clustering effect. On the basis of clustering,introduced the abnormal factors concept, ICAD algorithm marks outlier factor as a standard marker of abnormal data, after the integration of several clustering through the data sorted according to the number of abnormal marks , get abnormal data sets. Experiments illustrate, CO-aiNet clustering algorithm is betterthan similar algorithm, andit has a considerablyenhance in the accuracy and stability based on clustering fusion anomaly detection.

关 键 词：混沌理论 免疫聚类 异常检测 聚类融合 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]