Towards Elimination of Redundant and Well Known Patterns in Spatial Association Rule Mining

In this paper we present a new method for mining spatial association rules from geographic databases. On the contrary of most existing approaches that propose syntactic constraints to reduce the number of rules, we propose to use background geographic information extracted from geographic database schemas. In a first step we remove all well known dependences explicitly represented in geographic database schemas. In a second step we remove redundant frequent sets. Experiments show a very significant reduction of the number of rules when both well known dependences and redundant frequent sets are removed.