一种基于SVR节点数据预测改进算法

摘要

To solve the problem of unable to assess the status of the monitoring area after some nodes unavailable or failure,a data prediction algorithm for wireless sensor network is put forward. Neighbor nodes factors are joined together with Support Vector Regression.It adjusts the data by the correlation of neighbor nodes to improve the SVR algorithm. It can solve inaccurate prediction problem of the SVR algorithm in random incident. The simulation results show that it can effectively deal with the changes leaded by the accidental factors and has enhance the accuracy of prediction.%针对无线网络中部分传感器失效后无法进行有效或准确预测节点信息的问题,提出了一种面向无线传感网络的节点数据预测算法.该算法是在支持向量回归(SVR)算法基础上,引入了邻居节点影响因素(或称邻居节点数据的相关性),采用邻居节点相关性对其数据进行修正,从而实现了对SVR算法的改进,弥补了SVR算法在随机突发事件中预测不够准确的问题.经过仿真实验表明,该算法能够有效地应对突发或临时原因引起采集样本数据突然改变问题,预测更接近真实数据,准确性更高.