摘要
针对大区域无线传感网络流量特征维度较高,现有神经网络算法只能假设所有区域特征一致,导致一旦网络规模过大,会存在较大误差的问题,采用烟花算法优化粒子群算法的搜索能力,设计了一种烟花算法-粒子群优化(Fireworks Algorithm-Particle Swarm Optimization,FW-PSO)算法,提升了全局搜索能力及收敛速度,并将之应用于大区域无线传感网络流量异常检测。首先采用时间滑动窗口处理大区域无线传感网络数据流信息,通过正态Grubbs法则剔除数据中粗大误差值。然后引入主成分分析法对传感数据特征降维,分段提取有价值的特征。设计FW-PSO算法,提升粒子群算法的搜索能力,实现流量异常检测。实验结果表明,所提方法的无线传感网络流量异常检测率准确率平均为94.8%,训练及检测耗时平均值分别为3.75 s及0.25 s。
In view of the high dimension of traffic characteristics in large-area wireless sensor networks,the existing neural network algo-rithms can only assume that all areas have the same characteristics,which leads to the problem of large errors once the network scale is too large.The fireworks algorithm is used to optimize the search ability of particle swarm optimization,and the designed FW-PSO(Fire-works Algorithm-Particle Swarm Optimization)algorithm has improved global search ability and convergence speed,which can be used in large area wireless sensor network traffic anomaly detection.Firstly,the time sliding window is used to process the data flow informa-tion of the large area wireless sensor network,and the coarse error value in the data is eliminated by adopting the normal Grubbs rule.Then principal component analysis method is introduced to reduce the dimension of sensor data features and extract valuable features in segments.The FW-PSO algorithm is designed to improve the search capability of the particle swarm algorithm and realize traffic anoma-ly detection.The experimental results show that the average accuracy of the proposed method for traftic anomaly detection of wireless sensor network is 94.8%,and the average training and detection time are 3.75 s and 0.25 s respectively.
作者
张兵
卞利
ZHANG Bing;BIAN Li(Institute of Information Engineering,Suqian College,Suqian Jiangsu 223800,China;College of Computer Science and Communication Engineering,Jiangsu University,Zhenjiang Jiangsu 212013,China)
出处
《传感技术学报》
CAS
CSCD
北大核心
2023年第7期1116-1121,共6页
Chinese Journal of Sensors and Actuators
基金
江苏省科技计划项目(BE2021354)
宿迁市科技计划项目(L202109)。
