摘要
针对混沌时间序列的混沌性,提出一种改进的相空间重构方法——交集寻优法;针对传统的BP神经网络、RBF神经网络及AR模型对混沌时间序列预测效率和预测精度较低的缺点,提出两种不同的Hermite神经网络预测模型。以四阶蔡氏电路为模型,结合粒子群算法建立预测模型。仿真结果表明,利用交集寻优法进行相空间重构能很好地保留原系统的动力学特性,证实了该方法的有效性;Hermite神经网络较传统的预测模型精度更高,便于基于粒子群算法的Hermite神经网络预测方法的推广和应用。
For chaotic property of chaotic time series,we proposed an improved phase space reconstruction method — the intersection optimisation method. In view of the shortcomings of traditional BP neural network,RBF neural network and AR model in low prediction efficiency and accuracy on chaotic time series,we put forward two different Hermite neural network prediction models. Taking the fourth-order chua＇s circuit as the model we built the prediction model in combination with PSO algorithm. Simulation results indicated that to reconstruct phase space using intersection optimisation method could well keep the dynamics characteristic of original system,thus the effectiveness of the method was confirmed. Hermite neural network has higher prediction accuracy than traditional neural network,it is easy to promote and apply the PSO-based Hermite neural network prediction method.
出处
《计算机应用与软件》
CSCD
2016年第4期268-272,共5页
Computer Applications and Software
关键词
相空间重构
Hermite神经网络
粒子群算法
混沌时间序列预测
Phase space reconstruction
Hermite neural network
Particle swarm optimisation（PSO）
Chaotic time series prediction
