statistical functions as an auxiliary means in hydrological time series modeling using new data driven techniques

 

interesting in using of relatively new data driven modeling techniques (eg. neural networks or neurofuzzy) is being increased among water resources engineering researchers in recent years. one of difficult task in hydrological modeling using these new computational intelligence techniques is selecting appropriate number of model inputs. in this study, the ability of using auto-, cross-, and partial autocorrelation functions were evaluated for model input selection. the method was applied for sembrong dam s reservoir inflow data located in malaysia. the selected inputs were used to construct simple neural network and neuro-fuzzy models. results show using these statistical functions can reduce attempts in modeling hydrological time series using computational intelligence techniques.