Integration of an adaptive neuro-fuzzy inference system (ANFIS) model for groundwater level prediction utilizing feature engineering techniques in the training process

Groundwater is a vital resource in Erbil City, as it is used for various human activities, including urbanization and industrialization, which has resulted in the depletion of groundwater. This study employs Artificial Neural Networks (ANNs) and Fuzzy Inference System (FIS) clustering techniques to predict groundwater levels in the Erbil region, Iraq, with a primary focus on feature engineering. The Adaptive Neuro-Fuzzy Inference System (ANFIS) method is a hybrid technique that uses backpropagation and least squares estimation to create fuzzy membership functions and parameters. First, a detailed investigation was conducted to demonstrate the enhanced performance of clustered models that incorporate previous Groundwater Level (GWL) data for various well sites. Second, the authors aimed to validate the developed model using GWL data from six separate wells in two different aquifers across distinct zones. Finally, this study proposes a novel approach to predict the rate of change in GWL rather than the absolute groundwater level. The major findings indicate that the ANFIS method achieved a remarkably high level of accuracy in modelling GWL when feature engineering techniques were considered. Furthermore, this work is valuable for enhancing the understanding of groundwater response in the Erbil region and for proposing an efficient approach for managing groundwater resources