摘要
准确估计链路质量是提高无线传感器网络(WSN)协议性能的基础,针对目前WSN链路质量估计模型中存在的特征选择以及分类方法没有统一的标准、缺少公开数据集等问题,提出了一种基于支持向量机的WSN链路质量分类估计模型APSO-SVM,利用互信息系数选择模型输入特征,通过APSO粒子群优化算法获得模型的最优超参数,根据输入特征的特点对缺失值进行填充。在多种场景和不同的干扰及竞争条件下进行实验,实验结果表明,与现有算法相比,本文提出的APSO-SVM具有更好的性能,查准率和查全率分别达到97.21%和96.47%。
Accurate estimation of link quality is the basis for improving the performance of wireless sensor network protocols. In view of the problems existing in the current WSN link quality estimation model, such as data imbalance, feature selection and lack of uniform standards for classification methods, a WSN link quality classification estimation model APSO-SVM based on support vector machine is proposed. The mutual information coefficient is used to select the model input features, and the optimal parameters of the model are obtained through the APSO particle swarm optimiza-tion algorithm, fill in the missing values according to the characteristics of the input features. Ex-periments are carried out in various scenarios and under different interference and competition conditions. The experimental results show that the proposed APSO-SVM has better performance than the existing algorithms, with a precision rate of 97.21% and a recall rate of 96.47%.
出处
《建模与仿真》
2023年第3期2217-2228,共12页
Modeling and Simulation
