Published: Građevinar 70 (2018) 1
Paper type: Preliminary report
Download article (Croatian): PDF
Download article (English): PDF
Application of artificial neural networks for hydrological modelling in Karst
Abstract
The possibility of short-term water flow forecasting in a karst region is presented in this paper. Four state-of-the-art machine learning algorithms are used for the one day ahead forecasting: multi-layer perceptron neural network, radial basis function neural network, support vector machines for regression (SVR), and adaptive neuro fuzzy inference system (ANFIS). The results show that the ANFIS model outperforms other algorithms when the root mean square error and mean absolute error are used as quality indicators.
Keywordsartificial neural networks, SVR, ANFIS, rainfall-runoff ratio in karst areas