Intelligent learning algorithms integrated with feature engineering for sustainable groundwater salinization modelling: Eastern Province of Saudi Arabia

Abstract

Groundwater (GW) salinization refers to the natural or human-induced process of GW acquiring a higher level of salt content. While geological factors or proximity to the ocean can naturally cause this phenomenon, it can also be attributed to activities like irrigation, changes in land use, and the discharge of industrial or municipal waste. The following research employs the use of various feature extraction methods coupled with novel chemometrics approaches informed by machine learning (ML) techniques including; Gaussian Process Regression (GPR), Support Vector Regression (SVR), Regression tree (RT) and Robust linear regression (RLR). Based on the feature selection methods, the models were classified into three different combinations. The intelligent learning algorithms equally depict higher PCC values ranging from 0.935 to 1.00 in the training and 0.779 to 0.999 in the validation stage respectively, which depicts a higher relation between the experimental and simulated values. The performance results indicate that GPR-Comb-3 showed the highest performance in both the training and validation stages respectively. It is worth mentioning that even the RLR technique equally depicts exceptional prediction skills in both the training and validation steps. In conclusion, the outcomes of the current research depict the significance of these techniques in evaluating GW salinization.