摘要
数据融合是无线传感器网络中减少节点能量消耗的一个基本方法.在基于预测的时域数据融合中,通过对传感器节点采集的时间序列数据进行分析,建立能够反映时间序列中所包含的动态依存关系的数学模型,从而减少节点间冗余数据的传输.本文引入流行学习中局部线性重构的思想,结合改进的极限学习机(Extreme Learning Machine),提出KNN-PSOELM数据预测模型.首先运用K近邻的方法对输入样本点进行局部线性重构,然后采用粒子群优化算法(Particle Swarm Optimization)改进极限学习机回归方法产生最优的初始参数.优化之后的模型不仅使得原始非线性传感器数据具有线性的特征,而且避免由异常数据样本引起的病态隐层输出矩阵,提高了模型的预测精度和泛化能力.实验结果表明了算法的有效性.
Effective data fusion has been emerged as a basic approach in wireless sensor networks （ WSNs ） to reduce the energy consumption. In the prediction-based data fusion, through the analysis of the periodically sensed data from sensor nodes, a mathematical model reflecting the dynamic dependencies in the time domain data is established so as to reduce the transmission of redundant data between nodes. This paper learns the idea of manifold theory, combined K-nearest neighbor （ KNN ） techniques with Extreme Learning Machine （ ELM ） optimized by the Particle Swarm Optimization （ PSO ）, presenting a novel data prediction model called KNN-PSOELM. Considering the fact the sensor nodes data is nonlinear, the KNN algorithm is used to reconstruct the raw data. Then a hy-brid learning algorithm is proposed to overcome the drawbacks of ELM, which uses an improved PSO algorithm to select the best input weights and hidden biases. The proposed model not only makes the raw sensor data keep the characteristics of nonlinear, but also avoid the ill-conditioned hidden output matrix H raised by the randomly choosing the input weights and hidden biases. Experiment results have verified the efficiency and effectiveness of the proposed method.
出处
《小型微型计算机系统》
CSCD
北大核心
2016年第11期2478-2482,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61103175
61300104)资助
福建省科技创新平台建设基金项目(2009J10007)资助
福建省自然科学基金项目(2013J01230)资助
福建省杰出青年科学基金项目(2014J06017)资助
福建省高等学校新世纪优秀人才支持计划基金项目(JA13021)资助
关键词
无线传感器网络
粒子群优化
局部线性重构
极限学习机
wireless sensor networks
particle swarm optimization
local linear reconstruction
extreme learning machine
