Combining a neural network with deterministic chaos theory using phase space reconstruction for daily rainfall-runoff forecasting

Abstract

Chaotic analysis of hydrological series revealed the presence of chaotic structures. As such, a Chaotic Neural Network model was proposed for daily rainfall-runoff. The approach is based on the combination of the series generated by the reconstruction of the phase space according to the method of Takens, in an artificial neural network. The results are very encouraging and open the prospects for other intelligent hybrid models taking into account the long dependency and multiscale effect optimized by genetic algorithms.