Fazendo as melhores escolhas - um estudo sobre aprendizado de máquina e a utilização de foco de algoritmos em desenho de grafos

• Ano de defesa: 2015
• Autor(a) principal: Vieira, Raissa dos Santos
• Orientador(a): Nascimento, Hugo Alexandre Dantas do
• Banca de defesa: Nascimento, Hugo Alexandre Dantas do, Laureano, Gustavo Teodoro, Ribeiro, Marcela Xavier
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Goiás
• Programa de Pós-Graduação: Programa de Pós-graduação em Ciência da Computação (INF)
• Departamento: Instituto de Informática - INF (RG)
• País: Brasil
• Palavras-chave em Português: Desenho de grafos; Ações de usuários; Raciocínio baseado em casos
• Palavras-chave em Inglês: Graph drawing; User actions; Case-based reasoning
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://repositorio.bc.ufg.br/tede/handle/tede/4884

The graph drawing problem is to construct geometric representations of graphs in order to obtain good drawings that meet user expectations. The problem becomes complex because the concept of “good drawing” is subjective and relative, in other words, it can vary according to the preferences or needs of each user. This paper presents a literature review of approaches that explored to apply machine learning techniques for drawing graphs. Then, it proposes a framework to collect user actions from interactions with a graph drawing software and reuse them using case-based reasoning. The framework was tested with a database interaction involving focus an genetic algorithm. The goal was to determine whether the reuse of these actions made by users could lead to an effective strategy for improvement graph drawings. Experiments were performed with the database and algorithms, described along with the framework, in order to evaluate the proposal approach through the statistical analysis of results obtained. The analysis showed promising strategies, among them an algorithm that matches a preexisting genetic algorithm running on the entire graph drawing, and the combination of this genetic algorithm with a new algorithm that can produces better graph drawings. Such facts motivate further research in this area.