Extração de conhecimento em bases de dados espaciais: algoritmo CHSMST+

• Ano de defesa: 2014
• Autor(a) principal: Medeiros, Camila Alves de [UNESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Computação; Sistemas de informação geografica; Infra-estruturas de dados espaciais; Banco de dados; Análise espacial (Estatística); Geographic information systems
• Link de acesso: http://hdl.handle.net/11449/110986

The development of technologies for collecting spatial information has resulted in a large volume of stored data, which makes inappropriate the use of conventional data mining techniques for knowledge extraction in spatial databases, due to the high complexity of these data and its relationships. Therefore, several algorithms have been proposed, and the spatial clustering ones stand out due to their high applicability in many fields. However, these algorithms still need to overcome many challenges to reach satisfactory results in a timely manner. In this work, we present a new algorithm, namely CHSMST+, which works with spatial clustering considering both distance and similarity, allowing to correlate spatial and non-spatial attributes. These tasks are performed without input parameters and user interaction, eliminating the dependence of the user interpretation for cluster generation and enabling the achievement of cluster in a more efficient way, since the calculations performed by the algorithm are more accurate than visual analysis of them. Together with these techniques, we use a multithreading approach, which allowed an average reduction of 38,52% in processing time. The CHSMST+ algorithm was applied in spatial databases of health and environment, showing the ability to apply it in different contexts, which makes this work even more relevant