Using Boosted Regression Tree, Logistic Model Tree, and Random Forest Algorithms to Evaluate the Groundwater Potential

Groundwater is exploited uncontrollably due to population growth and industrialization in different parts of the world. The purpose of this study is to evaluate the groundwater potential by advanced machine learning algorithms using topographical, hydrological, environmental, and geological criteria. To do this, three advanced machine learning algorithms were used, including Boosted Regression Tree (BRT), Logistic Model Tree (LMT), and Random Forest (RF). Therefore, for implementation, geo-hydrological data of 37 groundwater wells in Birjand plain of South Khorasan province were collected and randomly selected in a ratio of 70 to 30 were divided into training and validation data sets. Finally, groundwater potential maps were prepared using BRT, LMT, and RF algorithms. In order to validate the groundwater potential prediction algorithms, the area under the curve (AUC) and the statistical criteria of positive predictive rate, negative predictive rate, sensitivity, specificity, and accuracy were used. The results showed that the LMT model (AUC = 0.865) has a better performance than the BRT and RF models in predicting the groundwater potential of the study area.