Flow Prediction Versus Flow Simulation Using Machine Learning Algorithms
- 1.4k Downloads
The paper deals with differences between two types of machine learning river flow modelling, i.e., their simulation and prediction. In this paper, “simulation” means a determination of river flows from only meteorological data. The second type of modelling, i.e., prediction, additionally includes preceding flows in the input data. Preceding flows are known at the time of making a prediction. For this reason, i.e., because less input data serve for the simulation, it is a more difficult task than the prediction, and its degree of precision is also usually lower. The authors focused on the improvement of flow simulation methodology, i.e., the determination of river flows only from climate data. Several machine learning models were tested for this purpose, and their results are compared in the paper with a conceptual hydrological model. Three options were evaluated in the paper for the improvement of the precision of the machine learning type of flows simulation: (1) the effect of the use of different types of models, (2) the impact from the expansion of input data utilizing feature engineering, and (3) improving the accuracy of the simulation by applying an ensemble paradigm. An increased degree of precision (approximately 12%) of the flow simulation was obtained after the incorporation of the above methodological enhancements to the computations (when compared to standard hydrological methods). The authors believe that the proposed methodology will be a promising alternative to the usual hydrological simulation, and it would be useful to test it in an extended study in which more streams would be evaluated.
KeywordsFlow simulation Flow prediction Data-driven methods
This work was supported by the Slovak Research and Development Agency under Contract No. APVV-15-0489 and by the Scientific Grant Agency of the Ministry of Education of the Slovak Republic and the Slovak Academy of Sciences, Grant No. 1/0665/15.
- 2.Papacharalampous, G.A., Tyralis, H., Koutsoyiannis, D.: Comparison of stochastic and machine learning methods for multi-step ahead forecasting of hydrological processes. Preprints 2017, 2017100133. https://doi.org/10.20944/preprints201710.0133.v1
- 5.Szalai, S., Spinoni, J., Galos, B., Bessenyei, M., Molar, P., Szentimrey, T.: Use of regional database for climate change and drought. In: 5th IDRC Davos 2014. Global Risk Forum GRF Davos (2014)Google Scholar
- 6.Viglione, A., Parajka, J.: TUWmodel: Lumped Hydrological Model for Education Purposes. R package version 0.1-8 (2016). https://CRAN.R-project.org/package=TUWmodel
- 11.Liaw, A., Wiener, M.: Classification and regression by randomForest. R News 2(3), 18–22 (2002)Google Scholar
- 12.R Core Team: R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria (2017). https://www.R-project.org/
- 14.Chen, T., Guestrin, C.: XGBoost: a scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 785–794. ACM (2016)Google Scholar
- 15.XGBoost Homepage. https://xgboost.readthedocs.io/en/latest/. Accessed 16 Mar 2018
- 16.Allaire, J.J., Chollet, F.: keras: R Interface to ‘Keras’. R package version 2.1.4 (2018). https://CRAN.R-project.org/package=keras