Avaliação de agrupamentos em mistura de variáveis

• Ano de defesa: 2013
• Autor(a) principal: Vidigal, Bruno Caetano
• Orientador(a): Cecon, Paulo Roberto
• Banca de defesa: Ferreira, Adésio
• 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: Análise de agrupamentos; Simulação; Algoritmos
• Palavras-chave em Inglês: Cluster analysis; Simulation; Algorithms
• Área do conhecimento CNPq: CNPQ::CIENCIAS AGRARIAS
• Link de acesso: http://locus.ufv.br/handle/123456789/4065

Cluster analysis is widely used in many research areas in order to recognize a standard structure of variability between individuals or objects studied, classifying them into homogeneous groups. However, the studies that are published, most of them deal only on numeric variables, excluding the analysis, the information contained in categorical variables. Thus, this study aims to evaluate some similarity measures and clustering algorithms in databases and also simulated on a case study in Genetics. The similarity measures evaluated were: euclidean, squared euclidean, mean euclidean, mahalanobis, manhattan, combined measures and gower. The hierarchical clustering algorithms are: nearest neighbor, furthest neighbor, UPGMA and Ward. The algorithms evaluated from the class of non-hierarchical are the kmeans and k-prototypes, which is an extension of the first. The results were compared and we concluded the non-hierarquical were better than hierarquical methods.