基于战争策略算法优化回声状态网络的时间序列预测

Optimizing time series prediction of echo state networks based on war strategy algorithm

为了解决回声状态网络(ESN)储备池参数难以确定的问题,提出一种基于战争策略优化算法(WSO)的回声状态网络模型(WSO-ESN).该模型利用战争策略优化算法中攻击和防御两种流行的战争策略更好地实现整个模型在全局探索和局部开发上的平衡,替换弱士兵策略提高其鲁棒性使WSO算法在确定ESN参数时更准确.此外,还引入了呈指数变化的权重更新机制提高算法的收敛速度进而更快地确定储备池参数.实验结果与粒子群优化算法(PSO)、蜣螂优化算法(DBO)、金豺优化算法(GJO)等对储备池参数优化方法进行比较.结果表明,基于战争策略优化算法的回声状态网络模型具有更快的训练速度和更高的预测精度.

In order to solve the problem of difficult to determine the parameters of the reserve pool of echo state network(ESN),this paper proposes an echo state network model(WSO-ESN)based on the war strategy optimization algorithm(WSO).The model utilizes two popular war strategies,attack and defense,in the war strategy optimization algorithm to better achieve a balance between global exploration and local exploitation of the whole model,and replaces the weak soldier strategy to improve its robustness so that the WSO algorithm will be more accurate in determining the ESN parameters.In addition,an exponentially varying weight updating mechanism is introduced to improve the convergence speed of the algorithm and thus determine the reserve pool parameters faster.The experimental results are compared with those of particle swarm optimization(PSO),dung beetle optimization(DBO),and golden jackal optimization(GJO)for optimizing the reserve pool parameters.The results show that the echo state network model based on war strategy optimization algorithm proposed in this paper has faster training speed and higher prediction accuracy.