Possibilities of using neural networks for data preprocessing in models predicting flash floods

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00020699%3A_____%2F21%3AN0000045" target="_blank" >RIV/00020699:_____/21:N0000045 - isvavai.cz</a>

Výsledek na webu

—

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Possibilities of using neural networks for data preprocessing in models predicting flash floods

Popis výsledku v původním jazyce

In the article are studied possibilities of preprocessing of the radar values using methods artificial intelligence (neural network). For this study were chosen 229 meteorological stations. The data from these stations are compared with mean values of radar data. The neural networks are trained on historical episodes (2016-2019) and whole model is tested on validation period, which it was chosen year 2020. The preprocessed data and were given to model for forecasting of flash flood danger and results of both inputs were compared. Preprocessed rainfall data significantly lowered number of fake alarms, but slightly increased number of missed dangerous events. Results of neural networks model were good enough for another continuation of this applications. Where should be find some problematic issues with the neural network application as preprocessing tool for this application.

Název v anglickém jazyce

Possibilities of using neural networks for data preprocessing in models predicting flash floods

Popis výsledku anglicky

In the article are studied possibilities of preprocessing of the radar values using methods artificial intelligence (neural network). For this study were chosen 229 meteorological stations. The data from these stations are compared with mean values of radar data. The neural networks are trained on historical episodes (2016-2019) and whole model is tested on validation period, which it was chosen year 2020. The preprocessed data and were given to model for forecasting of flash flood danger and results of both inputs were compared. Preprocessed rainfall data significantly lowered number of fake alarms, but slightly increased number of missed dangerous events. Results of neural networks model were good enough for another continuation of this applications. Where should be find some problematic issues with the neural network application as preprocessing tool for this application.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10501 - Hydrology

Projekt

<a href="/cs/project/VI20192021166" target="_blank" >VI20192021166: Hydrometeorologická rizika v České republice - změny rizik a zlepšení jejich predikcí</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2021

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

XXIX Danube Conference - XXIX Conference of the Danubian Countries on Hydrological Forecasting and Hydrological Bases of Water Management - Conference proceedings - Extended abstracts

ISBN

978-80-7653-017-1

ISSN

—

e-ISSN

—

Počet stran výsledku

2

Strana od-do

73-74

Název nakladatele

Czech Hydrometeorological Institute

Místo vydání

Praha

Místo konání akce

Brno

Datum konání akce

6. 9. 2021

Typ akce podle státní příslušnosti

EUR - Evropská akce

Kód UT WoS článku

—