Abstract:
Based on neural network theory and phase-space reconstruction theory, a prediction algorithm for chaotic time series of optimized Radial Basis Function(RBF) neural based on Differential Evolution(DE) is proposed. In order to get the optimal neural network predictive model, the center, width, and connection weights of RBF neural networks are optimized by the global search ability of DE. The availability of the prediction algorithm is proved by the simulation of three typical nonlinear systems. Compared with the forecasting results of RBF neural network, results show that the improved algorithm has better generalization ability and higher forecasting accuracy.
Key words:
chaotic time series,
prediction,
Radial Basis Function(RBF) neural network,
Differential Evolution(DE) algorithm,
phase-space reconstruction,
nonlinear system
摘要： 以神经网络和相空间重构相关理论为基础，提出一种基于差分进化(DE)优化径向基函数(RBP)神经网络的改进混沌时间序列预测算法。利用DE的全局搜索能力优化RBF神经网络基函数的中心、宽度以及网络的连接权值，以此获得最优的网络预测模型。将该预测算法应用于3种典型的非线性系统进行有效性验证，并与RBF神经网络预测模型的预测结果进行比较。仿真结果表明，改进算法的泛化能力优于RBF网络，同时可提高网络的预测精度。
关键词:
混沌时间序列,
预测,
径向基函数神经网络,
差分进化算法,
相空间重构,
非线性系统
CLC Number: