Application of multilayer perceptron (MLP) method for streamflow forecasting (case study: Upper Citarum River, Indonesia)

Abstract Flood forecasting is a critical component of flood early warning. The discharge that occurs one the parameters can be used as reference for predicting flooding. Various models based on physically or data-driven have been developed. One methods considered watersheds with limited physical data Artificial Neural Network (ANN). Furthermore, ANN method allows analysis process to completed in less time and fewer resources. employed this work multilayer perceptron (MLP). MLP model was developed study predict streamflow at Citarum river, particularly Dayeuhkolot hydrological station 2, 4, 6, 8, 10, 12, 24 hours ahead. Two input scenarios were modeling scene. First, form rainfall data. second regional Before coming hours, hyperparameters optimized using GridSearchCV method. model’s performance assessed RMSE, NSE, R 2 values. produced satisfactory results both when than 4 NSE value higher 0.9. Scenario produces slightly better prediction scenario 1. Based values, more 6 accurate results.