基于优化回声状态网络的混沌时间序列预测

A prediction for chaotic time series based on optimized echo state network

基于多种群遗传算法(multiple population genetic algorithm,MPGA)优化回声状态网络(echo state networks,ESN)的储备池参数设置方案,建立MPGA-ESN模型,并将其用于上证指数开盘价预测的仿真实验.通过与BP神经网络、Elman神经网络、PSO-ESN模型、GA-ESN模型的对比,发现MPGA-ESN模型的预测精度更佳.

A MPGA-ESN model is set up based on multiple population genetic algorithm （MPGA） to optimize the parameters of the ESN reserve pool, and this model is used in the simulation experiment of the Shanghai composite index opening price forecast. By comparing with the BP neural network, Elman neural network, PSO-ESN, GA-ESN model, it is found that MPGA-ESN model has better prediction accuracy.