abstract
- Machine Learning algorithms have been applied to a variety of problems in hydrology. In this work, the aim is to generate a model able to predict two values of a water stream: stage and discharge, for periods of time that could range several weeks. The input for the model are still images of the river. This paper analyzes features computed by hydrologists and use them to compare several machine learning models. The models tested are: Random Forest, Multilayer Perceptron, K-Nearest Neighbors and Support Vector Machine. The results show that is possible to generate a reasonably good model with all the features. It was also analyzed the selection of attributes with two methods. A simpler model, with a small decrease in accuracy was obtained by this means. The model was able to predict for longer periods of time than the ones reported previously. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.