Towards Elimination of Well Known Geographic Patterns in Spatial Association Rule Mining

Many spatial association rule mining algorithms have been developed to extract interesting patterns from large geographic databases. However, a large amount of knowledge explicitly represented in geographic database schemas has not been used to reduce the number of association rules. A significant number of well known dependences, explicitly represented by the database designer, are unnecessarily extracted by association rule mining algorithms. The result is the generation of hundreds or thousands of well known spatial association rules. This paper presents an approach for mining spatial association rules where both database and schema are considered. We propose the APRIORI-KC (a priori knowledge constraints) algorithm to eliminate all associations explicitly represented in geographic database schemas. Experiments show a very significant reduction of the number of rules and the elimination of well known rules