Mineração de regras de associação temporais envolvendo dados quantitativos contínuos

• Ano de defesa: 2020
• Autor(a) principal: João, Rafael Stoffalette
• Orientador(a): Ribeiro, Marcela Xavier
• Tipo de documento: Tese
• 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: Mineração de dados; Regras de associação; Temporalidade; Dados quantitativos contínuos
• Palavras-chave em Inglês: Data mining; Association rules; Temporality; uantitative continuous data
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://repositorio.ufscar.br/handle/20.500.14289/13877

The consideration of temporality in a explicit manner on the task of association rules mining that involves continuous quantitative data is one approach that aims to contribute to the field of study of knowledge discovery in databases. The construction of temporal intervals from attributes of a data set also provides to the method to identify binary relations, which these intervals may have. This work describes the development of a new method, named ART-Q, for the task of mining temporal association rules which involve continuous quantitative data. The temporality is assumed, in the present work, in its explicit form, not only by data sequencing. The patterns that allow the rules construction are made of binary relations from Allen’s intervalar algebra in the temporal intervals that describe the continuous quantitative attribute’s behavior of interest. The method has proven being able to reveal implicit information in different contexts databases. The results are demonstrated by temporal intervals of interest of the attributes and their algebric relations, patterns and temporal association rules. The work demonstrates that the method ART-Q contributes to the evolution of the literature with the definition and search of a new kind of pattern, more complex than those present in the studies. By the consideration of this kind of patterns association rules are constructed semantically involving a large amount of information among the implication of the rule.