Mineração de dados usando programação genética

• Autor(a) principal: Duarte, Mariana de Luna Freire
• Data de Publicação: 2012
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UFPB
• Texto Completo: https://repositorio.ufpb.br/jspui/handle/tede/6094
• topic: Banco de dados Geográficos; Mineração de Dados Espaciais; Algoritmos Evolucionários; Programação Genética; Geographic Databases; Spatial Data Mining; Evolutionary Algorithms; Genetic Programming; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO

Data mining has become an important activity for decision-making in large and small companies since it allows the extraction of relevant and non-trivial information so that corrections and adjustment in administrative and economic strategies could be selected. Consequently, an increase in the geographical data storage is seen in such a way that conventional data mining cannot carry out the extraction of knowledge from a high dimension database. According to the current literature, there are few tools capable of extracting knowledge from geographical data, mainly if the database is made of conventional (numeral and textual) and geographical (point, line and polygon) data. The aim of this study is to present a new algorithm for spatial data mining DMGP using the two types of data to carry out the information extraction from a determined base. This algorithm is based on the DMGeo algorithm which also seeks to extract knowledge from the two types of data. These algorithms are based on Genetic Programming and were developed to obtain classification rules of patterns existing in the numeral and geographical attributes. To obtain a better performance for the DMGeo, the use of meta-heuristic GRASP and ILS in the performance of DMGP algorithm was proposed to improve the individuals from the generated population . GRASP and ILS were used to generate the initial population and disturb some individuals aiming at finding better solutions.