|
چکیده
|
The presence of chromium (Cr) in groundwater poses a significant threat to human health. However, the lack of testing in many wells suggests that the severity of this issue may be underestimated. In this study, various predictive models, including soft computing techniques such as gene expression programming (GEP), artificial neural networks (ANN), multivariate adaptive regression splines (MARS), and the M5 Tree model, along with Random Forest (RF) and Multiple Linear Regression (MLR), were employed to estimate geogenic Cr concentrations in groundwater based on geological and geochemical parameters in northeastern Iran. A dataset of 676 Cr concentration measurements was used to train and evaluate the models. Among the methods tested, ANN demonstrated the highest predictive accuracy, followed closely by RF, which provided competitive results. GEP and MARS also showed reasonable performance, while MLR exhibited the weakest accuracy, highlighting the limitations of linear models in addressing complex geochemical processes. The ANN model identified over 600,000 individuals in the central and western regions of the study area as being at significant risk of geogenic Cr contamination in groundwater. The findings underscore the potential of advanced predictive models in groundwater quality management and their applicability in other regions with similar challenges.
|