Melhoria do tratamento de obstáculos na abordagem de agrupamento de dados espaciais SWMU clustering

• Ano de defesa: 2020
• Autor(a) principal: Gallo, Gabriel Passatuto
• Orientador(a): Ciferri, Ricardo Rodrigues
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de São Carlos; Câmpus São Carlos
• Programa de Pós-Graduação: Programa de Pós-Graduação em Ciência da Computação - PPGCC
• Palavras-chave em Português: Banco de dados; Dados espaciais; Mineração de dados; Agrupamento de dados; Agricultura de precisão; Zonas de manejo
• Palavras-chave em Inglês: Database; Spatial data; Data mining; Data clustering; Precision agriculture; Management zones
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::METODOLOGIA E TECNICAS DA COMPUTACAO::BANCO DE DADOS
• Link de acesso: https://repositorio.ufscar.br/handle/20.500.14289/13671

The technological has been improved considerably in recent years, providing the great benefits to several areas of application. Among these areas, agriculture had a great boost, enabling the increasing of the production and at the same time reducing costs and environmental impacts through crop management techniques, thus in this sense practicing the concepts of Precision Agriculture (AP). One of the methods used in PA is to design the planted area in smaller plots with similar values of soil and plant attributes, known as management zones or differentiated management units (UGDs). In this way, spatial data clustering algorithms are used to create UGD maps, in which they depict soil variability. Spatial Ward’s Management Units Clustering (SWMU Clustering) is an approach to spatial data clustering that enables the design of UGDs in AP. Its main advantage over related approaches is the significant reduction of stratification in clusters, obtaining maps of UGDs that are easily interpretable by the end user. This Master’s research investigated how to improve the management of spatial obstacles performed by the SWMU Clustering approach. In this sense, two new strategies were proposed: Replacement Strategy for the Set of Internal Samples to Obstacles and Buffer Strategy. These strategies were compared to the original strategy of the SWMU Clustering approach, showing that the Buffer strategy generated the best results. In addition, as a result of this research, an web application was developed for the SWMU Clustering approach, making it available as a service so that the end user can interact with the SWMU Clustering ap, from sending the input data until the visualization of the UGD results.