Level Prediction of Rivers In The Hydrographic Region of The Paraguay River Using Machine Learning Algorithms

The Pantanal is part of the Paraguay River Hydrographic Region (RH-Paraguay) and is characterized by the flooding of a portion of its area during certain times of the year. In this context, this study investigates the application of Machine Learning (ML) techniques for predicting river levels in the RH-Paraguay. Thus, a dataset with daily level values is used, with a sample of three stations selected. Next, the optimal hyperparameters for Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Bidirectional Long Short-Term Memory (BiLSTM) networks were explored, followed by network training and evaluation. Subsequently, the best-performing models for each algorithm were selected and compared with the Regression technique currently in use. The results show that all three models present improvements over the current model. The model using the GRU algorithm stood out for having the lowest error rates and being 23.84% more accurate than the Regression model, while LSTM and BiLSTM are, respectively, 18.09% and 19.16% more accurate than the Regression model. The LSTM and BiLSTM models approximate the actual values more closely at the peaks of maximum and minimum levels.