o:4171 Machine Learning Regression Models Analysis: Piezometric water level prediction - case study en Abstract: Recent development of artificial intelligence, machine learning and deep learning, in particular, resulted in the increase in the use of data-based models in various fields; among others, in the field of dam safety. Neural networks are the most frequently used machine learning technique which has been applied to various problems. Other machine learning techniques are used for the analysis and interpretation of dam structural behaviour. In this paper, an analysis is conducted exhibiting how novel machine learning techniques can be used for piezometric water level prediction. Results from different techniques are presented and discussed. At the same time, the performance of the previously developed neural network model is analysed with the extended dataset, since additional measurements have been collected in the meantime. Although only one representative piezometer is considered, the proposed methodology may be generally applicable. Finally, some recommendations are given on how predictive models that are very similar at first glance may differ by additional analyses. neural networks, machine learning, deep learning, dam safety 1552100 978-86-81037-71-3 2024-03-15T11:21:29.452Z 44 no 46 Vukašin Ćirović Institut za vodoprivredu "Jaroslav Černi"; Jaroslav Černi Water Institute 0009-0006-1980-5986 Nikola Milivojević Institut za vodoprivredu "Jaroslav Černi"; Jaroslav Černi Water Institute person 0000-0001-5330-1219 Vesna Ranković Faculty of Engineering, University of Kragujevac, Kragujevac, Serbia person 0000-0002-5445-9971 application/pdf 2035828 https://unilib.phaidrabg.rs/o:4171 no yes 1 70 92000004 71A04 First Serbian International Conference on Applied Artificial Inteligence (SICAAI), Kragujevac, Serbia, May 19-20, 2022. 1 4 University of Kragujevac, Serbia 2022