基于DBN实现QSSVM模型的WSN数据异常检测  

Wireless Sensor Network Anomaly Detection of QSSVM Model Based on DBN

作　　者：谷军闪 Gu Junshan(School of Information Engineering,Henan Vocational University of Science and Technology,Zhoukou Henan 466000,China)

机构地区：[1]河南科技职业大学信息工程学院,河南周口466000

出　　处：《山西电子技术》2024年第2期9-11,共3页Shanxi Electronic Technology

基　　金：河南省高等学校重点科研项目(22A470005)。

摘　　要：无线传感器网络(Wireless Sensor Network,WSN)数据异常处理效率保障了现在智能的高效率运行。在分析1/4超球面支持向量机(Quarter-Sphere support vector machines,QSSVM)测试模型的基础上,进行深度信念网络(Deep Belief Network,DBN)构建实现,设计了一种可以实现在线测试功能的异常检测算法。研究结果表明:随着窗口大小的增加,所需要的计算时间增多。QSSVM在窗口开始扩大时便产生变化,主要表现在准确度的持续提高。QSSVM检测性能随着样本维度不断升高得到较大提升,相反K-means的检测性能却有降低趋势。采用QSSVM算法处理560维HAR数据时,测试结果显示检测率高达94.16%。该研究能够满足大规模高维传感器的数据处理需求,具有很高的应用价值。Wireless Sensor Network(WSN)data exception processing efficiency guarantees the high efficiency of intelligent operation.On the basis of analyzing the test model of Quarter-Sphere support vector machines(QSSVM),the Deep Belief Network(DBN)is constructed and implemented.This paper designs an anomaly detection algorithm which can realize online test function.The results show that the computing time increases with the increase of window size.QSSVM changes as the window begins to expand,mainly in the form of continuous improvement in accuracy.The detection performance of QSSVM is greatly improved with the continuous increase of sample dimensions,while the detection performance of K-means has a decreasing trend.When QSSVM algorithm is used to process 560 dimensional HAR data,the test results show that the detection rate is as high as 94.16%.The research can meet the data processing requirements of large-scale high dimensional sensors and has high application value.

关 键 词：传感器网络 数据异常 深度信念网络 超球面支持向量机 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]