基于SVDD算法的高维非线性传感器数据流异常点检测  

Detection of Anomaly Points in High Dimensional Nonlinear Sensor Data Flow Based on SVDD Algorithm

作　　者：高峰 GAO Feng(Fuzhou University of Technology,Fuzhou 350000,China)

机构地区：[1]福州理工学院,福州350000

出　　处：《安阳工学院学报》2024年第2期65-69,共5页Journal of Anyang Institute of Technology

基　　金：福建省教育厅项目“基于深度学习的推荐技术优化研究及应用”(JAT191007)。

摘　　要：传感器数据通常具有高维特征,且在实际工业环境中可能存在复杂的非线性关系。如何有效处理高维度、非线性和复杂的数据特征,是当前研究的难点问题。为此本研究基于支持向量数据描述算法,针对高维非线性传感器数据流进行异常点检测。采用生成对抗网络提取高维非线性数据特征,通过主成分分析方法对提取得到的特征进行降维处理,以减少数据维度。使用经过降维处理的数据训练SVDD模型,通过求解对偶问题,得到的支持向量的系数和阈值等参数,以确定异常检测的决策边界,根据决策边界实现异常点检测。通过实验验证可知,所提方法的异常检出率较高,误报率较低,得出SVDD算法在高维非线性传感器数据流中异常点检测具有有效性,证明了其在实际工程应用中的潜在价值。Sensor data typically has high-dimensional features and may exhibit complex nonlinear relationships in actual industrial environments.Handling high-dimensional,nonlinear,and complex data features is a challenging issue in current research.This study is based on the Support Vector Data Description(SVDD)algorithm for anomaly detection in high-dimensional nonlinear sensor data streams.Generative adversarial networks are used to extract high-dimensional nonlinear data features,and principal component analysis is employed to reduce the dimensionality of the extracted features.The SVDD model is trained using dimensionality-reduced data,the dual problem is solved,and parameters such as support vector coefficients and thresholds are obtained to determine the decision boundary for anomaly detection.Anomaly point detection is achieved Based on the decision boundary.Experimental verification show that the proposed method has a high anomaly detection rate and a low false alarm rate.The study concludes that the SVDD algorithm is effective in detecting outliers in high-dimensional nonlinear sensor data streams,showcasing its potential value in practical engineering applications.

关 键 词：SVDD算法 传感器数据流 异常点 生成对抗网络 主成分分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]