Application of machine learning and deep learning for modeling a water basin

Detalhes bibliográficos
Autor(a) principal: de Azevedo Silva, Vinícius
Data de Publicação: 2023
Outros Autores: de Feiras Souza, Rafael, Peixoto Oliveira, Mateus, Lledo dos Santos, Francisco
Tipo de documento: Artigo
Idioma: por
Título da fonte: P@ranoá
Texto Completo: https://periodicos.unb.br/index.php/paranoa/article/view/47488
Resumo: Economic development associated with aspects such as social inequality is a process that leads to the deterioration of environmental resources. It is essential to improve technologies to combat the effects of natural disasters, such as enhancing the processes of rainfall-to-runoff transformation and modeling hydrological phenomena in river basins. The objective of this scientific communication was to obtain hydrological behavior predictions in the Bananal River Basin using machine learning techniques. The gathered information was subjected to the following models: deep learning artificial neural networks (InceptionTime, Resnet, FCN, and ResCNN) and machine learning models (linear regression, decision tree, and support-vector machine), as well as gradient boosting, to analyze consistent historical data and compare precipitation patterns and streamflow (rainfall-runoff model) in the watercourse. Daily average time series data from 2002 to 2014 were used for training to establish optimized behavior for 2015. It was concluded that the architecture with the highest performance, based on the coefficient of determination (R²), was InceptionTime, with an RMSE of 4.62 and MAE of 2.36. The decision tree model achieved the best overall performance with an R² of 0.73.
id UNB-28_b050487a04cd330794881ba0dbada832
oai_identifier_str oai:ojs.pkp.sfu.ca:article/47488
network_acronym_str UNB-28
network_name_str P@ranoá
repository_id_str
spelling Application of machine learning and deep learning for modeling a water basinAplicación de aprendizaje máquino y aprendizaje profundo para modelar una cuenca de aguaAplicação de machine learning e deep learning para modelagem de uma bacia hidrográficaDeep Learning, inundaçõeschuva-vazãomodelagemInceptionTimeinundacionesagua de lluviaFlujo de diseñomodeladSimulación ComputacionalFloodrainfallDesign flowComputational modellingEconomic development associated with aspects such as social inequality is a process that leads to the deterioration of environmental resources. It is essential to improve technologies to combat the effects of natural disasters, such as enhancing the processes of rainfall-to-runoff transformation and modeling hydrological phenomena in river basins. The objective of this scientific communication was to obtain hydrological behavior predictions in the Bananal River Basin using machine learning techniques. The gathered information was subjected to the following models: deep learning artificial neural networks (InceptionTime, Resnet, FCN, and ResCNN) and machine learning models (linear regression, decision tree, and support-vector machine), as well as gradient boosting, to analyze consistent historical data and compare precipitation patterns and streamflow (rainfall-runoff model) in the watercourse. Daily average time series data from 2002 to 2014 were used for training to establish optimized behavior for 2015. It was concluded that the architecture with the highest performance, based on the coefficient of determination (R²), was InceptionTime, with an RMSE of 4.62 and MAE of 2.36. The decision tree model achieved the best overall performance with an R² of 0.73.El desarrollo económico asociado con aspectos como la desigualdad social es un proceso que provoca la deterioración de los recursos ambientales. Es urgente mejorar las tecnologías para combatir los efectos de los desastres naturales, como el mejoramiento de los procesos de lluvia en caudal y la modelización de los fenómenos hidrológicos en cuencas hidrográficas. El objetivo fue obtener, mediante técnicas de aprendizaje automático, pronósticos del comportamiento hidrológico en la Cuenca del Río Bananal. Esta información recopilada se sometió a los siguientes modelos: redes neuronales artificiales de aprendizaje profundo (Deep Learning - InceptionTime, Resnet, FCN y ResCNN) y modelos de aprendizaje automático (Machine Learning - regresión lineal, árbol de decisión y máquina de vectores de soporte), así como impulso de gradiente para generar series históricas consistentes y comparar los patrones de precipitación y flujo (modelo de lluvia-caudal) del curso del agua. Se utilizaron las series temporales promedio diarias de 2002 a 2014 para el entrenamiento, con el fin de establecer un comportamiento optimizado para 2015. Se pudo concluir que el mejor rendimiento, según el coeficiente R², fue InceptionTime, con un RMSE de 4,62 y un MAE de 2,36. El modelo de árbol de decisión obtuvo el mejor R² de 0,73.O desenvolvimento econômico associado a aspectos como a desigualdade social são processos que provocam a deterioração de recursos ambientais. São emergenciais o aperfeiçoamento das tecnologias para combater os efeitos dos desastres naturais, como a melhoria dos processos de transformação de chuva em vazão e da modelagem dos fenômenos hidrológicos em bacias hidrográficas. O objetivo desta comunicação científica foi obter, com o auxílio de técnicas de machine learning, previsões do comportamento hidrológico na Bacia do Rio Bananal. Essas informações reunidas foram submetidas aos seguintes modelos: redes neurais artificiais de aprendizado profundo (Deep Learning - InceptionTime, Resnet, FCN e ResCNN) e de aprendizado de máquina (Machine Learning - regressão linear, árvore de decisão e support-vector machine), e gradient boosting para levantar séries históricas consistentes e comparar os padrões de precipitação e da vazão (modelo chuva-vazão) do curso hídrico. Foram utilizadas as séries temporais médias diárias de 2002 a 2014 para treinamento a fim de estabelecer um comportamento otimizado para 2015. Foi possível concluir que o coeficiente R² com a melhor performance foi da arquitetura InceptionTime, com RMSE de 4.62, MAE de 2.36. O resultado com amior desempenho foi a árvore de decisão com R2 de 0.73.Programa de Pós-graduação em Arquitetura e Urbanismo, Faculdade de Arquitetura e Urbanismo2023-09-06info:eu-repo/semantics/articleinfo:eu-repo/semantics/publishedVersionArtigo avaliado pelos paresapplication/pdfhttps://periodicos.unb.br/index.php/paranoa/article/view/4748810.18830/issn.1679-0944.n34.2023.20Paranoá; Vol. 16 No. 34 (2023): January/June Edition; 1-21Paranoá; v. 16 n. 34 (2023): Fluxo Contínuo - Janeiro/Junho; 1-211679-09441677-7395reponame:P@ranoáinstname:Universidade de Brasília (UnB)instacron:UNBporhttps://periodicos.unb.br/index.php/paranoa/article/view/47488/38484Copyright (c) 2023 Paranoáhttps://creativecommons.org/licenses/by/4.0info:eu-repo/semantics/openAccessde Azevedo Silva, Viníciusde Feiras Souza, RafaelPeixoto Oliveira, MateusLledo dos Santos, Francisco2024-02-08T15:00:47Zoai:ojs.pkp.sfu.ca:article/47488Revistahttps://periodicos.unb.br/index.php/paranoaPUBhttps://periodicos.unb.br/index.php/paranoa/oaiparanoa@unb.br1679-09441677-7395opendoar:2024-02-08T15:00:47P@ranoá - Universidade de Brasília (UnB)false
dc.title.none.fl_str_mv Application of machine learning and deep learning for modeling a water basin
Aplicación de aprendizaje máquino y aprendizaje profundo para modelar una cuenca de agua
Aplicação de machine learning e deep learning para modelagem de uma bacia hidrográfica
title Application of machine learning and deep learning for modeling a water basin
spellingShingle Application of machine learning and deep learning for modeling a water basin
de Azevedo Silva, Vinícius
Deep Learning
, inundações
chuva-vazão
modelagem
InceptionTime
inundaciones
agua de lluvia
Flujo de diseño
modelad
Simulación Computacional
Flood
rainfall
Design flow
Computational modelling
title_short Application of machine learning and deep learning for modeling a water basin
title_full Application of machine learning and deep learning for modeling a water basin
title_fullStr Application of machine learning and deep learning for modeling a water basin
title_full_unstemmed Application of machine learning and deep learning for modeling a water basin
title_sort Application of machine learning and deep learning for modeling a water basin
author de Azevedo Silva, Vinícius
author_facet de Azevedo Silva, Vinícius
de Feiras Souza, Rafael
Peixoto Oliveira, Mateus
Lledo dos Santos, Francisco
author_role author
author2 de Feiras Souza, Rafael
Peixoto Oliveira, Mateus
Lledo dos Santos, Francisco
author2_role author
author
author
dc.contributor.author.fl_str_mv de Azevedo Silva, Vinícius
de Feiras Souza, Rafael
Peixoto Oliveira, Mateus
Lledo dos Santos, Francisco
dc.subject.por.fl_str_mv Deep Learning
, inundações
chuva-vazão
modelagem
InceptionTime
inundaciones
agua de lluvia
Flujo de diseño
modelad
Simulación Computacional
Flood
rainfall
Design flow
Computational modelling
topic Deep Learning
, inundações
chuva-vazão
modelagem
InceptionTime
inundaciones
agua de lluvia
Flujo de diseño
modelad
Simulación Computacional
Flood
rainfall
Design flow
Computational modelling
description Economic development associated with aspects such as social inequality is a process that leads to the deterioration of environmental resources. It is essential to improve technologies to combat the effects of natural disasters, such as enhancing the processes of rainfall-to-runoff transformation and modeling hydrological phenomena in river basins. The objective of this scientific communication was to obtain hydrological behavior predictions in the Bananal River Basin using machine learning techniques. The gathered information was subjected to the following models: deep learning artificial neural networks (InceptionTime, Resnet, FCN, and ResCNN) and machine learning models (linear regression, decision tree, and support-vector machine), as well as gradient boosting, to analyze consistent historical data and compare precipitation patterns and streamflow (rainfall-runoff model) in the watercourse. Daily average time series data from 2002 to 2014 were used for training to establish optimized behavior for 2015. It was concluded that the architecture with the highest performance, based on the coefficient of determination (R²), was InceptionTime, with an RMSE of 4.62 and MAE of 2.36. The decision tree model achieved the best overall performance with an R² of 0.73.
publishDate 2023
dc.date.none.fl_str_mv 2023-09-06
dc.type.driver.fl_str_mv info:eu-repo/semantics/article
info:eu-repo/semantics/publishedVersion
Artigo avaliado pelos pares
format article
status_str publishedVersion
dc.identifier.uri.fl_str_mv https://periodicos.unb.br/index.php/paranoa/article/view/47488
10.18830/issn.1679-0944.n34.2023.20
url https://periodicos.unb.br/index.php/paranoa/article/view/47488
identifier_str_mv 10.18830/issn.1679-0944.n34.2023.20
dc.language.iso.fl_str_mv por
language por
dc.relation.none.fl_str_mv https://periodicos.unb.br/index.php/paranoa/article/view/47488/38484
dc.rights.driver.fl_str_mv Copyright (c) 2023 Paranoá
https://creativecommons.org/licenses/by/4.0
info:eu-repo/semantics/openAccess
rights_invalid_str_mv Copyright (c) 2023 Paranoá
https://creativecommons.org/licenses/by/4.0
eu_rights_str_mv openAccess
dc.format.none.fl_str_mv application/pdf
dc.publisher.none.fl_str_mv Programa de Pós-graduação em Arquitetura e Urbanismo, Faculdade de Arquitetura e Urbanismo
publisher.none.fl_str_mv Programa de Pós-graduação em Arquitetura e Urbanismo, Faculdade de Arquitetura e Urbanismo
dc.source.none.fl_str_mv Paranoá; Vol. 16 No. 34 (2023): January/June Edition; 1-21
Paranoá; v. 16 n. 34 (2023): Fluxo Contínuo - Janeiro/Junho; 1-21
1679-0944
1677-7395
reponame:P@ranoá
instname:Universidade de Brasília (UnB)
instacron:UNB
instname_str Universidade de Brasília (UnB)
instacron_str UNB
institution UNB
reponame_str P@ranoá
collection P@ranoá
repository.name.fl_str_mv P@ranoá - Universidade de Brasília (UnB)
repository.mail.fl_str_mv paranoa@unb.br
_version_ 1797066980618928128

Registros relacionados