Efficacy of neural network and genetic algorithm techniques in simulating spatio-temporal fluctuations of groundwater

摘要: Groundwater modelling has emerged as a powerful tool to develop sustainable management plan for efficient groundwater utilization and protection of this vital resource. This study deals with the development five hybrid artificial neural network (ANN) models their critical assessment simulating spatio-temporal fluctuations in an alluvial aquifer system. Unlike past studies, study, all relevant input variables having significant influence on have been considered, ANN technique [ANN-cum-Genetic Algorithm (GA)] used simulate levels at 17 sites over area. The parameters were optimized using GA optimization technique. predictive ability developed each was evaluated six goodness-of-fit criteria graphical indicators, together adequate uncertainty analyses. analysis results revealed that multilayer perceptron Levenberg–Marquardt model is most predicting monthly almost sites, while radial basis function least efficient. found be superior commonly trial-and-error method determining optimal architecture internal parameters. Of statistics only root-mean-squared error, r2 Nash–Sutcliffe efficiency more useful assessing performance models. It can concluded approach effectively basin or subbasin scales. Copyright © 2014 John Wiley & Sons, Ltd.