WSN中利用XGBoost和加权自适应HFLMS的数据约减组合预测方法

针对无线传感器网络(WSN)中能量、带宽和内存等各种资源的限制问题,提出了一种XGBoost结合加权自适应分层分数最小均方误差(hierarchical fractional least-mean-square,HFLMS)的数据约减组合预测方法。首先,利用XGBoost方法对损失函数进行了二阶的泰勒展开,权衡模型的复杂度和损失函数的下降速度,实现了资源限制的稳定预测;然后提出自适应HFLMS滤波器实现WSN数据约简的传输,并基于误差估计来预测所感测的数据,有效降低了WSN中的能量约束;最后,利用两个评估参数(能量和预测误差)来验证所提组合预测方法的性能。实验结果表明,相比没有预测、近似最速下降算法和分层最小均方滤波技术,提出的预测方法获得的预测结果更好。

Data reduction combination prediction method using XGBoost and weighted adaptive HFLMS in WSN

Aiming at the limitation of various resources such as energy,bandwidth and memory in a wireless sensor network(WSN),this paper proposed a data reduction combination prediction method based on XGBoost and hierarchical fractional least-mean-square(HFLMS).Firstly,it used the XGBoost method to perform a second-order Taylor expansion of the loss function,which balanced the complexity of the model and the decline rate of the loss function,and achieved the stable prediction of the resource limit.Then,it employed the proposed HFLMS for data reduction in WSN,and used error estimation to predict the measured data,which would reduce the energy constraints in WSN.Finally,it used the two evaluation parameters(energy and prediction error)to evaluate the performance of the proposed prediction method.The experimental results demonstrate that the proposed prediction method is better than that without prediction,the approximate steepest descent algorithm and the layered minimum mean square filtering technology.