Assessment and prediction of water quality indices by machine learning-genetic algorithm and response surface methodology

Conventional techniques for determining water adequacy are generally expensive because they take into account a variety of factors. Thus, the problem of water management in agriculture may benefit from the development of precise and reliable models. The goal of this research is to forecast parameters related to agriculture and human use in the southeast Mediterranean Sea (North Africa), including dam water quality index (WQI), soluble sodium percentage (SSP), sodium absorption rate (SAR), and Kelly rate (KR). In this study, Central Composite Design (CCD) is combined with Response Surface Methodology (RSM), and a Genetic Algorithm (GA)-based Artificial Neural Network (ANN) was implemented using data from Principal Component Analysis (PCA), which accounted for 65.20% of the total data set. The results show that Back Propagation Neural Networks (BPNNs) predictive models outperform the RSM approach for predicting water quality indexes. The results show that the accuracy of the RSM models (R² = 97.30%, 98.90%, 97.88%, and 97.27%) is lower than that of the ANN predictions for WQI, SAR, KR, and SSP. Furthermore, ANN improves water quality management and allows agriculturalists to control water quality indices. This research could be useful for developers looking for accurate water quality data to help them develop water supply management strategies.