Implementation of rough sets theory for knowledge acquisition and construction of knowledge base for generalization of geographic information

Abstract

Generalization of geographic information was always based on the knowledge and experience of cartographer who performed it. Progressive technological changes, especially the change in data storage from analog to electronic devices (usually in the form of databases), changed a lot in generalization process. Though, the knowledge needed for it is still crucial, it is stored in different way as well. Today, knowledge bases are built based on years of cartographer’s experience. Their goal is to enable, at least partially, automation of the generalization process.
This knowledge base for modern spatial data may be constituted, among others, by information about attributes of objects which are significant from the point of view of generalization process as well as the generalization rules based on those attributes. Contemporary spatial databases include a number of attributes and other information (as geometry and topology) which can be used to calculate other databases. Therefore, the challenge is to bring to light information crucial in generalization process. This paper presents the use of rough sets concept, specifically reducts, for this goal. The reduct is a subset of the original set of attributes which allows to make decisions (in this case, the decision about generalization of objects) as good as based on original attributes.
Knowledge about the attributes significant in generalization process is derived, in this way, from already existing data. Selected attributes can be used for defining the generalization rules. Thanks to the use only of the attributes constituting reducts, the construction of rules becomes easier and the rules themselves, as well as generalization process, are more transparent.