| |
|
|
| Registros recuperados : 105 | |
| Registros recuperados : 105 | |
|
|
|
 | Acesso ao texto completo restrito à biblioteca da Embrapa Agricultura Digital. Para informações adicionais entre em contato com cnptia.biblioteca@embrapa.br. |
Registro Completo
|
Biblioteca(s): |
Embrapa Agricultura Digital. |
|
Data corrente: |
07/12/2018 |
|
Data da última atualização: |
07/01/2020 |
|
Tipo da produção científica: |
Resumo em Anais de Congresso |
|
Autoria: |
FARHATE, C. V. V.; SOUZA, Z. M. de; OLIVEIRA, S. R. de M.; LOVERA, L. H.; OLIVEIRA, I. N. de; GUIMARÃES, E. M. |
|
Afiliação: |
CAMILA VIANA VIEIRA FARHATE, Feagri/Unicamp; ZIGOMAR MENEZES DE SOUZA, Feagri/Unicamp; STANLEY ROBSON DE MEDEIROS OLIVEIRA, CNPTIA; LENON HERIQUE LOVERA, Feagri/Unicamp; INGRID NEHMI DE OLIVEIRA, Feagri/Unicamp; EURIANA MARIA GUIMARÃES, Feagri/Unicamp. |
|
Título: |
Data mining techniques for classification of soil CO2 emission. |
|
Ano de publicação: |
2018 |
|
Fonte/Imprenta: |
In: WORLD CONGRESS OF SOIL SCIENCE, 21., 2018, Rio de Janeiro. Soil science: beyond food and fuel: abstracts. Viçosa, MG: SBCS, 2018. |
|
Páginas: |
Não paginado. |
|
Idioma: |
Inglês |
|
Notas: |
WCSS 2018. |
|
Conteúdo: |
A high priority objective currently in the scope of carbon cycle science is to understand the spatial and temporal controls involved in CO2 dynamics in terrestrial ecosystems. However, estimates of CO2 emissions from soil to the atmosphere through production systems are difficult and complex due to the diversity of agricultural practices in large areas and significant variations in both soil and climate. In contrast, data mining is a promising alternative to predict soil CO2 emission from correlated variables. Thus, our objective was to construct a model using data mining techniques, such as selection of attributes and induction of decision trees to predict different levels of CO2 emissions in the soil. The original data set was composed of 23 attributes (22 predictive attributes and one response variable). The response variable refers to the emission of CO2 from the soil as the target of the classification. Due to the large number of attributes, a procedure for selecting attributes was conducted to remove those of low correlation to the response variable. For this purpose, we assessed four approaches to attribute selection: no attribute selection, correlation-based attribute selection (CFS), Chi-square method (χ2), and Wrapper method. For data classification, we used the binary decision tree induction technique on Weka 3.6 software. Our results demonstrated that the data mining techniques allowed the development of an efficient model to classify soil CO2 emission using the Wrapper method of attribute selection as well as algorithm C4.5 for induction of the decision tree. Wrapper method selected an efficient subset for soil respiration prediction with only five attributes, with the following influence order on the determination of soil CO2 emission: soil temperature> rainfall> macroporosity> soil moisture> potential acidity. The attributes selected through the Wrapper method have high coherence with literature data regarding both the selected attributes and the decision tree rules. MenosA high priority objective currently in the scope of carbon cycle science is to understand the spatial and temporal controls involved in CO2 dynamics in terrestrial ecosystems. However, estimates of CO2 emissions from soil to the atmosphere through production systems are difficult and complex due to the diversity of agricultural practices in large areas and significant variations in both soil and climate. In contrast, data mining is a promising alternative to predict soil CO2 emission from correlated variables. Thus, our objective was to construct a model using data mining techniques, such as selection of attributes and induction of decision trees to predict different levels of CO2 emissions in the soil. The original data set was composed of 23 attributes (22 predictive attributes and one response variable). The response variable refers to the emission of CO2 from the soil as the target of the classification. Due to the large number of attributes, a procedure for selecting attributes was conducted to remove those of low correlation to the response variable. For this purpose, we assessed four approaches to attribute selection: no attribute selection, correlation-based attribute selection (CFS), Chi-square method (χ2), and Wrapper method. For data classification, we used the binary decision tree induction technique on Weka 3.6 software. Our results demonstrated that the data mining techniques allowed the development of an efficient model to classify soil CO2 emission using... Mostrar Tudo |
|
Palavras-Chave: |
Árvore de decisão; Data mining; Decision tree; Emissão de gás carbônico; Mineração de dados; Selection of attributes; Soil attributes. |
|
Categoria do assunto: |
X Pesquisa, Tecnologia e Engenharia |
|
Marc: |
LEADER 02929nam a2200277 a 4500
001 2100970
005 2020-01-07
008 2018 bl uuuu u00u1 u #d
100 1 $aFARHATE, C. V. V.
245 $aData mining techniques for classification of soil CO2 emission.$h[electronic resource]
260 $aIn: WORLD CONGRESS OF SOIL SCIENCE, 21., 2018, Rio de Janeiro. Soil science: beyond food and fuel: abstracts. Viçosa, MG: SBCS$c2018
300 $aNão paginado.
500 $aWCSS 2018.
520 $aA high priority objective currently in the scope of carbon cycle science is to understand the spatial and temporal controls involved in CO2 dynamics in terrestrial ecosystems. However, estimates of CO2 emissions from soil to the atmosphere through production systems are difficult and complex due to the diversity of agricultural practices in large areas and significant variations in both soil and climate. In contrast, data mining is a promising alternative to predict soil CO2 emission from correlated variables. Thus, our objective was to construct a model using data mining techniques, such as selection of attributes and induction of decision trees to predict different levels of CO2 emissions in the soil. The original data set was composed of 23 attributes (22 predictive attributes and one response variable). The response variable refers to the emission of CO2 from the soil as the target of the classification. Due to the large number of attributes, a procedure for selecting attributes was conducted to remove those of low correlation to the response variable. For this purpose, we assessed four approaches to attribute selection: no attribute selection, correlation-based attribute selection (CFS), Chi-square method (χ2), and Wrapper method. For data classification, we used the binary decision tree induction technique on Weka 3.6 software. Our results demonstrated that the data mining techniques allowed the development of an efficient model to classify soil CO2 emission using the Wrapper method of attribute selection as well as algorithm C4.5 for induction of the decision tree. Wrapper method selected an efficient subset for soil respiration prediction with only five attributes, with the following influence order on the determination of soil CO2 emission: soil temperature> rainfall> macroporosity> soil moisture> potential acidity. The attributes selected through the Wrapper method have high coherence with literature data regarding both the selected attributes and the decision tree rules.
653 $aÁrvore de decisão
653 $aData mining
653 $aDecision tree
653 $aEmissão de gás carbônico
653 $aMineração de dados
653 $aSelection of attributes
653 $aSoil attributes
700 1 $aSOUZA, Z. M. de
700 1 $aOLIVEIRA, S. R. de M.
700 1 $aLOVERA, L. H.
700 1 $aOLIVEIRA, I. N. de
700 1 $aGUIMARÃES, E. M.
Download
Esconder MarcMostrar Marc Completo |
|
Registro original: |
Embrapa Agricultura Digital (CNPTIA) |
|
|
Biblioteca |
ID |
Origem |
Tipo/Formato |
Classificação |
Cutter |
Registro |
Volume |
Status |
Fechar
|
| Nenhum registro encontrado para a expressão de busca informada. |
|
|
