Nature-inspired approaches to mining trend patterns in spatial databases

Abstract

Large repositories of spatial data have been formed in various applications such as Geographic Information Systems (GIS), environmental studies, banking, etc. The increasing demand for knowledge residing inside these data-bases has attracted much attention to the field of Spatial Data Mining. Due to the common complexity and huge size of spatial databases the aspect of efficiency is of the main concerns in spatial knowledge discovery algorithms. In this paper, we introduce two novel nature-inspired algorithms for efficient discovery of spatial trends, as one of the most valuable patterns in spatial databases. The algorithms are developed using ant colony optimization and evolutionary search. We empirically study and compare the efficiency of the proposed algorithms on a real banking spatial database. The experimental results clearly confirm the improvement in performance and effectiveness of the discovery process compared to the previously proposed methods. © Springer-Verlag Berlin Heidelberg 2006.