Groundwater Level Prediction Using a Multiple Objective Genetic Algorithm-Grey Relational Analysis Based Weighted Ensemble of ANFIS Models

Predicting groundwater levels is critical for ensuring sustainable use of an aquifer’s limited reserves and developing a useful abstraction management strategy. The purpose this study was to assess the predictive accuracy estimation capability various models based on Adaptive Neuro Fuzzy Inference System (ANFIS). These included Differential Evolution-ANFIS (DE-ANFIS), Particle Swarm Optimization-ANFIS (PSO-ANFIS), traditional Hybrid Algorithm tuned ANFIS (HA-ANFIS) one- multi-week forward forecast at three observation wells. Model-independent partial autocorrelation functions followed by frequentist lasso regression-based feature selection approaches were used recognize appropriate input variables prediction models. performances evaluated using statistical performance evaluation indexes. results revealed that optimized performed equally well in predicting one-week-ahead wells when set indexes used. For improving accuracy, weighted-average ensemble proposed, which weights individual calculated Multiple Objective Genetic (MOGA). MOGA accounts benefits (higher values indicate better model performance) cost (smaller test dataset. Grey relational analysis select best solution from feasible solutions produced MOGA. A MOGA-based ranking superiority DE-ANFIS (weight = 0.827), HA-ANFIS 0.524), 0.697) GT8194046, GT8194048, GT8194049, respectively. Shannon’s entropy-based decision theory utilized rank result indicated outperformed all (ranking value 0.987, 0.985, 0.995 respectively). worst performers PSO-ANFIS 0.845), 0.819), 0.900) generalization proposed modelling approach forecasting 2-, 4-, 6-, 8-weeks ahead data GT8194046. confirmed useability higher horizons. demonstrated may be successfully predict multi-week-ahead levels, utilizing previous lagged as inputs.