Prediction Of Future Hydrological Droughts With Tree-Based Algorithms

This study used tree-based machine learning algorithms such as RF and GBM to predict short-term hydrological droughts. First, 3 and 12-month SRI values were calculated for hydrological droughts. In constructing the models, the SRI data is divided into 80% training and 20% testing. In addition, past SRI values were presented as inputs to the models and future SRI (t+1) values were forecasted. The study tested the potential of tree-based algorithms in predicting hydrological droughts according to various statistical indicators.