Mineração de regras de associação em dados georreferenciados / Mining of association rules in geographic data




Geographic information systems are used to store, manipulate, and analyze spatial data and its descriptive aspects. Spatial data analysis can be done by searching association rules that describe relationships between the data. However, georeferenced data present topological relations unknown to traditional mining association rule algorithms. To solve this problem, Koperski and Han (1995) and Malerba et al. (2001) proposed a topological relation extraction process integrated to a mining association rule algorithm. This process requires all data to be translated as logical predicates. As an alternative to this solution, this work proposes to break down this process by pre-processing the georeferenced data to find topological relations, then executing traditional mining association rule algorithms. The resulting rules must present descriptive characteristics of the data and topological relations. To reach this objective, a process of rule extraction in georeferenced data was specified, in addition to a pre-processing module implementation. This module was evaluated by using a case study that uses a geographic information system of the city of Jaboticabal, in the context of urban planning. The generated rules were analyzed by a specialist using the measures of support and confidence. In addition, an analysis regarding the processing time and memory consumption was provided to find the topological relations, which shows that it is possible to extract the patterns with the proposed process and module.


geographic data mineração de dados regras de associação banco de dados espaciais association rules data mining

Documentos Relacionados