@PhDThesis{Pessoa:2014:PrEvSe,
               author = "Pessoa, Alex Sandro Aguiar",
                title = "Predi{\c{c}}{\~a}o de eventos severos em sa{\'{\i}}das de 
                         modelos meteorol{\'o}gicos utilizando a teoria dos conjuntos 
                         aproximativos e metaheur{\'{\i}}sticas para redu{\c{c}}{\~a}o 
                         de atributos",
               school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
                 year = "2014",
              address = "S{\~a}o Jos{\'e} dos Campos",
                month = "2014-11-14",
             keywords = "teoria dos conjuntos aproximativos, metaheur{\'{\i}}sticas, 
                         eventos severos, rough set theory, metaheuristics, severe 
                         events.",
             abstract = "A Teoria dos Conjuntos Aproximativos (TCA) {\'e} um paradigma 
                         para tratamento de informa{\c{c}}{\~o}es incertas e imprecisas 
                         proposta no in{\'{\i}}cio dos anos 80 e vem se difundindo nas 
                         {\'u}ltimas duas d{\'e}cadas gra{\c{c}}as ao aumento das 
                         capacidades de processamento e armazenamento de dados. Um ponto 
                         central na TCA {\'e} a obten{\c{c}}{\~a}o de conjuntos 
                         reduzidos de atributos conhecidos como redu{\c{c}}{\~o}es, as 
                         quais reduzem a dimensionalidade da classifica{\c{c}}{\~a}o. 
                         Entretanto, a obten{\c{c}}{\~a}o de redu{\c{c}}{\~o}es a 
                         partir do conjunto completo de atributos possui alta complexidade 
                         computacional, recorrendo-se ent{\~a}o ao uso de 
                         metaheur{\'{\i}}sticas. Nesta tese, objetiva-se identificar 
                         padr{\~o}es associados {\`a} ocorr{\^e}ncia de eventos 
                         convectivos severos em sa{\'{\i}}das de modelos num{\'e}ricos 
                         de previs{\~a}o de tempo utilizando-se TCA. Estes padr{\~o}es 
                         s{\~a}o constitu{\'{\i}}dos por um conjunto selecionado de 
                         vari{\'a}veis meteorol{\'o}gicas e s{\~a}o encontrados a partir 
                         de um conjunto de eventos convectivos conhecidos, os quais foram 
                         identificados por meio da densidade de ocorr{\^e}ncia de 
                         descargas el{\'e}tricas nuvem-solo. A aplica{\c{c}}{\~a}o de 
                         metaheur{\'{\i}}sticas espec{\'{\i}}ficas otimiza a 
                         identifica{\c{c}}{\~a}o desses padr{\~o}es no escopo da TCA e 
                         permite gerar classificadores que possam detectar a 
                         poss{\'{\i}}vel ocorr{\^e}ncia de eventos convectivos em 
                         previs{\~o}es meteorol{\'o}gicas. Isso auxiliaria a 
                         previs{\~a}o operacional de tempo, dada a defici{\^e}ncia que os 
                         modelos meteorol{\'o}gicos tem em simular a g{\^e}nese e 
                         evolu{\c{c}}{\~a}o de eventos convectivos devida a 
                         limita{\c{c}}{\~o}es de resolu{\c{c}}{\~a}o espacial e {\`a} 
                         necessidade de se aprimorar a microf{\'{\i}}sica correspondente 
                         nesses modelos. ABSTRACT: The Rough Set Theory (RST) is a standard 
                         proposed to deal with uncertain, incomplete or vague information 
                         that was proposed in the early 80s. The use of RST has been 
                         spreading over the last two decades thanks to increase of data 
                         processing and storage capabilities. A fundamental point of RST is 
                         the calculation of reduced sets of attributes known as reducts, 
                         which allow to reduce the classification dimensionality. However, 
                         the calculation of reducts from the complete set of attributes 
                         presents high algorithmic complexity demanding the use of 
                         metaheuristics. The aim of this thesis is to identify patterns 
                         associated to the occurrence of severe convective events from the 
                         output of weather forecast numerical models using RST. These 
                         patterns are composed of a selected set of meteorological 
                         variables and are found using a set of known convective events, 
                         which were identified using the density of occurrence of 
                         cloud-to-ground electrical discharges. The application of specific 
                         metaheuristics optimizes the identification of such patterns in 
                         the scope of RST, and allows to derive classifiers able to detect 
                         the possible occurrence of convective events in weather forecasts. 
                         This approach would help the operational weather forecasting 
                         considering that meteorological models have poor performance to 
                         simulate the genesis and evolution of convective events due to 
                         spatial resolution limitations and to the need of improving the 
                         corresponding microphysics in such models.",
            committee = "Sandri, Sandra Aparecida (presidente) and Stephany, Stephan 
                         (orientador) and Dutra, Luciano Vieira and Ambrizzi, T{\'e}rcio 
                         and Forster, Carlos Henrique Quartucci",
           copyholder = "SID/SCD",
         englishtitle = "Prediction of severe convective events from weather model output 
                         using the rough set theory and metaheuristics for attribute 
                         reduction.",
             language = "pt",
                pages = "146",
                  ibi = "8JMKD3MGP3W34P/3HFJU3S",
                  url = "http://urlib.net/ibi/8JMKD3MGP3W34P/3HFJU3S",
           targetfile = "publicacao.pdf",
        urlaccessdate = "27 abr. 2024"
}