Machine learning-based modeling of suitable dam sites in Northern Ghana

ABSTRACT There is a crucial need for water resources in arid regions. The focus of this study is to examine areas suitable for siting dams using machine learning (ML) techniques. The study was conducted by applying the random forest (RF) and support vector machine (SVM) algorithms to predict regions suitable for siting a dam by utilizing factors that influence the process. The results of the study suggest that 54.43% of the study area was found suitable in the RF model, while 49.72% was found viable with the SVM model, which represents a moderate, high, and very high-suitability class in those areas. The ML algorithm was able to predict areas suitable for siting a dam with a root mean square error, R2, and minimal absolute error of 47.84, 0.53, 0.64, and 43.55, 0.49, 0.61 for the RF and SVM models, respectively. Also, it was found that the area under the receiver operating characteristic curve for the RF and SVM models was 0.827 and 0.790, respectively. The findings of the study can be applied to enhance the decision-making process regarding dam construction. This study offers a solid framework for pinpointing ideal dam locations by utilizing advanced ML methods.