一种基于G(1,1)改进模型的WSN流量异常检测方法

An Abnormity Detection Method of WSN Data Flow Based on Grey Prediction Theory

提出了一种基于灰系统预测理论的无线传感器网络数据流异常检测方法,并使用组合优化策略进行了改进,用轮转迭代法求得组合优化参数.实验结果表明:经过3次迭代,该方法即可较标准G(1,1)模型降低59.5%的误差,其能耗低于时间序列分析常用算法ARMA.该方法既可以在传感器节点使用,也可以在sink节点使用,具有所需原始数据少、建模快、实时性和适用性强、扩展性好等特点,弥补了传感器网络节点资源受限的不足.

An abnormity detection method of WSN data flow based on Grey System Prediction theory is presented in this paper. This method modified and optimized the basic grey system method G（1,1） with adapting different λ, parameter. Tests and simulation results show that the energy consumption with this method decreases greatly and obviously than ARMA, and only through 3 times iterations the prediction error decreases 59.5. This method can meet with the resource limitation of WSN.