Echo State Network Optimization Based on Improved Fireworks Algorithm

First, the standard firework algorithm (Firework Algorithm, FWA) is improved, and the initial firework population is generated by an adversarial learning strategy. At the same time, according to the displacement method of the explosion spark in the original FWA, the explosion spark generation method and an inertia weight factor in the enhanced firework algorithm are introduced. Then the improved FWA algorithm WOFWA (Weight Opposition Firework Algorithm, WOFWA) is combined with Echo State Networks (ESN) to construct a WOFWA-ESN model. Four classical complex nonlinear benchmark functions are selected as test functions to verify the superiority of the WOFWA algorithm, and the improved FWA-ESN model is used to predict three time series data sets. The experimental results show that the accuracy of the WOFWA-ESN prediction model proposed in this paper is significantly improved.