Prediction of Groundwater Salinization Using Particle Swarm Optimization for Neural Network Training

Monitoring groundwater quality is a costly and time-consuming process. The use of machine learning models has proven to be a suitable alternative for predicting groundwater quality indicators. In this work, an artificial neural network model has been trained via particle swarm optimization (PSO) with hydrochemical data collected from a coastal aquifer in Tunisia. The validity of the PSO-trained model is evaluated based on different performance indicators, demonstrating higher accuracy than the model trained with the traditional gradient descent method concerning all the evaluation metrics. These results are consistent with the well-known ability of these methods to perform more effective searching in the solution space.