基于D-S证据理论的无监督异常检测算法  

Unsupervised anomaly detection algorithm based on D-S evidence theory

作　　者：衷卫声[1] 吴自望 张强[1] ZHONG Weisheng;WU Ziwang;ZHANG Qiang(School of Advanced Manufacturing,Nanchang University,Nanchang 330031,China;School of Information Engineering,Nanchang University,Nanchang 330031,China)

机构地区：[1]南昌大学先进制造学院,江西南昌330031 [2]南昌大学信息工程学院,江西南昌330031

出　　处：《南昌大学学报（工科版）》2024年第2期255-261,共7页Journal of Nanchang University(Engineering & Technology)

基　　金：国家自然科学基金资助项目(62161022)。

摘　　要：在实际应用中,当数据集缺少真实标签或正常点数据量不足时,可能导致一分类支持向量机处于无监督情况。此外,当训练集中包含异常数据时,一分类支持向量机生成的决策边界将偏斜至异常数据区域。上述问题降低了异常数据的检测率,并导致分类器的性能变差。为了解决上述问题,基于K近邻算法将数据集分为可疑正常点数据集与可疑离群点数据集。其中,可疑正常点数据集用于一分类支持向量机训练与建模,对于可疑离群点数据集则采用D-S证据理论来识别其中的正常数据。实验结果表明:基于D-S理论的无监督异常检测算法可以有效地分离正常点与异常点,该算法在整体数据集上A_(uc)均值达到了0.83,且在可疑离群点数据集上A_(uc)均值达到了0.883。In practical applications,when the dataset is unlabeled or the amount of normal point data is insufficient usually leads to one-class support vector machine(OCSVM)operating in an unsupervised manner.In addition,when the training set includes anomalies,the decision boundary formed by OCSVM will skew toward to the anomalies.The above problems may undermine the detection rate of anomalies and result in poor performance of the classifier.In order to solve the above problems,we divided the dataset into suspicious normal point dataset and suspicious outlier dataset based on KNN algorithm.The suspicious normal point dataset was used for OCSVM training and modeling,and for the suspicious outlier dataset,the D-S evidence theory was utilized to identify the normal data of them.The experiment results showed the DS-SVM algorithm can effectively separate the normal points and the anomalies,the mean A_(uc) value of algorithm was 0.83 on the overall dataset,and 0.883 in the suspicious outlier dataset.

关 键 词：离群点检测 一分类支持向量机 DEMPSTER-SHAFER证据理论 无监督学习 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]