Filtering frequent geographic patterns with qualitative spatial reasoning.

Abstract

In frequent geographic pattern mining a large amount of patterns can be non-novel and non-interesting. This problem has been addressed recently, and background knowledge is used to reduce well known geographic patterns. However, a large amount of meaningless patterns which is independent of domain knowledge is still extracted from geographic data. Therefore, this paper proposes a method for filtering specific types of meaningless spatial patterns using qualitative spatial reasoning. We proof a significant reduction of the number of frequent patterns, which is also shown with experiments performed on real data. These experiments even show a reduction in computational time.