Classification and projection of spatial association rules

This paper proposes an approach allowing the definition of association rules relative to a future date from sets of rules that are relative to previous dates. The produced rules concern the future land use in urban areas. The suggested approach allows the integration of variable data into existing techniques of spatial data mining. The used process is based on a meta-rules generation in order to classify produced rules according to their temporal evolution. Subsequently, the technique of least squares is used to estimate the future values of rule confidence according to which applicable rules to a future date will be selected. A prototype and an experiment on a spatial database taken at various dates gave encouraging results.