Author(s): M. Siek; D. P. Solomatine
Linked Author(s):
Keywords: Chaos theory; Ensembles; Neural network; Ocean surge; North Se
Abstract: A problem of forecasting ocean surge based on high-frequency data on surge is considered. A technique based on optimized model ensembles in high-dimensional phase space is presented. Time-delayed phase space is built using the principles of chaos theory. The forecasts are made in this space by the adaptive local data-driven multi-models based on the neighbors of the current vector. The method is tested to make forecasts of storm surge for the North Sea. The results show that the multi-model ensemble model has a higher accuracy compared to a standard chaotic model or a single data-driven model.
Year: 2011