Application of machine learning to groundwater spring potential mapping using averaging, bagging, and boosting techniques

Abstract Determining groundwater potential is vital for resource management. This study aims to present a comparative analysis of three widely used ensemble techniques (averaging, bagging, and boosting) in spring mapping. Firstly, 12 spring-related factors total 79 locations were collected as the dataset. Secondly, typical models adopted predict potential, namely, Bayesian model averaging (BMA), random forest (RF), gradient boosting decision tree (GBDT). The area under receiver operating characteristics curve (AUC) four statistical indexes (accuracy, sensitivity, specificity, root mean square error (RMSE)) estimate model's accuracy. results indicate that had good predictive performance AUC values GBDT, RF, BMA 0.88, 0.84, 0.78, respectively. Furthermore, GBDT best (accuracy = 0.89, sensitivity 0.91, specificity 0.87, RMSE 0.33) terms indexes, followed by RF 0.83, 0.36) 0.76, 0.65, 0.49). research can provide effective guidance using mapping future.