基于人工鱼群算法优化神经网络的WSN数据融合

无线传感网(wireless sensor network， WSN)通常节点众多、数据冗余度高，传统的基于随机权值和阈值的前馈反向传播神经网络(back propagation neural network， BPNN)数据融合方法易陷入局部极值，导致融合结果准确性差。提出一种优化神经网络的权值和阈值进而改善WSN数据融合质量的方法-人工鱼群算法前馈反向传播(artificial fish swarm algorithm back propagation， AFSABP)神经网络数据融合。仿真和对比实验结果表明，改进的鱼群算法在收敛速度和寻优精度上都有明显提升，改进后的人工鱼群BP算法数据融合方法相较于传统BP数据融合方法，可减少3.06%的相对误差和3.74%的均方根误差。

WSN data fusion based on neural network optimized by artificial fish swarm algorithm

Wireless sensor network (WSN) usually holds a large number of nodes with high data redundancy. The traditional data fusion methods based on BP neural network which use random weights and thresholds easily fall into local extremum and then derive some inaccurate fusion results. A method is proposed to improve the quality of WSN data fusion by optimizing the weights and thresholds of neural networks. The artificial fish swarm algorithm BP (AFSABP) neural network data fusion. The results of simulation and experiments show that the improved algorithm has significant improvement in convergence speed and optimization accuracy, and the AFSABP method using the improved fish swarm algorithm can reduce the relative error 3.06% and the root-mean-square error 3.74% compared to the existing WSN neural network data fusion method.