Criação de um ambiente computacional para detecção de outliers e preenchimento de falhas em dados meteorológicos

• Ano de defesa: 2015
• Autor(a) principal: Ventura, Thiago Meirelles
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Mato Grosso; Brasil; Instituto de Física (IF); UFMT CUC - Cuiabá; Programa de Pós-Graduação em Física Ambiental
• Palavras-chave em Português: Tratamento de dados; Dados ambientais; Inteligência artificial; Framework; CNPQ::CIENCIAS EXATAS E DA TERRA::FISICA; Data correction; Environmental data; Artificial intelligence
• Link de acesso: http://ri.ufmt.br/handle/1/2186

In order to study the environment meteorological data series must be analyzed. However, these data series may contain errors, because of electronic failures, animal action or weather phenomena, among other factors. These failures can result in missing data or outliers, causing difficulties in the data analysis. Therefore, it is important to detect the outliers in the data series and fill in the missing data. This work presents a computational environment that will enable the correction of environmental data. In order to achieve this, three new methods were created in this work: one for gap filling and two for outlier detection. In addition, three other methods were obtained from other studies and were implemented together with the new methods in a single framework. These methods use techniques from the area of artificial intelligence and statistics, which often requires a deep study in order to apply them. However, the developed framework enables the application of these methods, only demanding the configuration of some parameters. Thus, the framework allows the development of applications with functionalities of gap filling and outlier detection. To demonstrate the applicability of these methods a web-based application was developed integrated with the framework. Besides, tests were carried out to verify the performance of each method created compared with those obtained from other studies. It is expected that this structure will increase the quality of data series, assisting in several scientific researches.