Multidimensional Nonlinearity Time Series Forecasting Based on Multi-reservoir Echo State Network
摘要
Echo state networks (ESNs) are known for their simple structure and excellent forecasting ability. In this chapter, we investigate the nonlinear prediction capabilities of ESNs, analyze their structure and principles, and propose a new prediction model based on the output mode, using multiple reservoirs. The proposed model is then tested using Mackey-Glass and Lorenz chaotic systems with different dimensions. Our results demonstrate that the multi-reservoir ESN model can accurately predict longer time series with high precision.