Mapeamento de áreas potenciais a implantação de aterro sanitário em Guarapuava-PR, com uso de redes neurais artificiais

• Ano de defesa: 2009
• Autor(a) principal: Antonio, Janaina Natali
• Orientador(a): Ribeiro, Selma Regina Aranha
• Banca de defesa: Nunes, João Osvaldo Rodrigues , Mathias, Ivo Mario
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: UNIVERSIDADE ESTADUAL DE PONTA GROSSA
• Programa de Pós-Graduação: Programa de Pós Graduação Mestrado em Gestão do Território
• Departamento: Gestão do Território : Sociedade e Natureza
• País: BR
• Palavras-chave em Português: aterro sanitário; resíduos sólidos domiciliares; redes neurais artificiais; Guarapuava-PR
• Palavras-chave em Inglês: sanitary embankment; solid waste; artificial neural networks; Guarapuava-PR
• Área do conhecimento CNPq: CNPQ::CIENCIAS HUMANAS::GEOGRAFIA
• Link de acesso: http://tede2.uepg.br/jspui/handle/prefix/523

The large production of solid waste and its wrong final destination in the environment have become a problem, especially in urban centers. For what effective measures can be taken, it is necessary to know the local reality, thus to be able act directly in the base of this question. The current system of managing solid residues in the municipality of Guarapuava-PR has as final destination the open atmosphere in "garbage", which causes a number of detriment to the environment and the people who live nearby. Considering this situation, the municipal administration has as proposal the construction of the one sanitary embankment. For the implementation of this must be considered technical criteria, politicians and social, for select an appropriate area. This work aims to carry through the mapping of areas with potential for deployment of sanitary embankment in Guarapuava-PR, using the methodology based on classification by Artificial Neural Networks (ANN). For classification were performed a series of tests with variations in the number of layers of input, and the parameters of the RNA, thus different results were obtained in the output layer. The best results were obtained with the architecture consists of 5 layers of input and 2 neurons in the hidden layer and change the variables of threshold training 0.8000, 0.1000 learning rate, dynamic training of 0.8000, mean square error of 0.0500 and number of iterations of 2000, is considered the ideal architecture for this type of classification.