Discriminação de populações com diferentes graus de similaridade por redes neurais artificiais

• Ano de defesa: 2009
• Autor(a) principal: Pereira, Tiago Martins
• Orientador(a): Cruz, Cosme Damião
• Banca de defesa: Peternelli, Luiz Alexandre , Carneiro, Pedro Crescêncio Souza
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Viçosa
• Programa de Pós-Graduação: Mestrado em Estatística Aplicada e Biometria
• Departamento: Estatística Aplicada e Biometria
• País: BR
• Palavras-chave em Português: Classificação; Análise discriminante; Redes neurais artificiais
• Palavras-chave em Inglês: Classification; Discriminant analysis; Artificial neural networks
• Área do conhecimento CNPq: CNPQ::CIENCIAS AGRARIAS
• Link de acesso: http://locus.ufv.br/handle/123456789/4026

The correct classification of individuals in pre-established groups has become of great importance in breeding. The multivariate statistical techniques commonly used in this type of problem are the discriminant functions of Fisher and Anderson, which are used to allocate an initially unknown individual in one of the g populations or pre-defined groups. In recent decades a new computing paradigm, artificial neural networks, has come along to solve various problems of Statistics, such as grouping of similar individuals, time series forecasting and also of particular interest, the problem of classification. The objective of this research was to conduct a simulation study in order to compare the discriminant functions of Fisher and Anderson and neural networks. We evaluated the number of incorrect classifications of individuals known to belong to different populations with different levels of dissimilarity measured by the Mahalanobis distance. Simulations were conducted using the software Genes (Cruz, 2006). Although Artificial Neural Networks presented a rate of incorrect classification of individuals rejected for being considered ambiguous as to its discriminatory characteristics, it proved to be a promising technique, since it presented a lower number of incorrect classifications of individuals when compared to the discriminant functions.