Application of statistics and machine learning methods for the analysis of hydrological data

Poplave mogu uzrokovati značajne gospodarske štete i ugroziti ljudske živote. U ovom radu napravljena je eksploratorna analiza podataka primjenom metoda statistike i strojnog učenja kako bi kasnije pomogla domenskim stručnjacima u upravljanju rizikom od poplava. Izdvojeni su nezavisni vodni valovi metodom godišnjih maksimuma, izračunati su dodatni indikatori na temelju izvornih podataka te su vodni valovi zatim grupirani prema obliku (na temelju interpolacije) i prema dodatnim indikatorima u po tri klastera koristeći particijsko grupiranje. Na kraju su predstavljene i uspoređene dobivene grupe klastera.Floods can cause significant economic damage and endanger human lives. In this study, an exploratory data analysis was conducted using statistical methods and machine learning to later assist domain experts in flood risk management. Independent water waves were extracted using the annual maxima method, additional indicators were calculated based on the original data, and the water waves were then grouped according to shape (based on interpolation) and additional indicators into three clusters using partition clustering. Finally, the obtained cluster groups were presented and compared