Abstract:
Several popular methods exist to characterize and analyze point pattern concentrations or “hot spots”. However, these methods (choropleth maps, clustering, spatial autocorrelation, etc.) do not provide any spatial context or additional information regarding features influencing the pattern, nor do they have the robustness to be trained in one location and applied in a second location. In addition, commonly used evaluation criteria are subjective and qualitative. This research will explore spatial analysis in feature space – an N-dimensional computational space where spatial entities are defined by their proximity to surrounding features, rather than their spatial coordinates - to further develop a multiple-reference spatial similarity model capable of eliciting significant knowledge to the structure of spatial point patterns. This is achieved by combining a model of spatial similarity with an exhaustive search optimization algorithm and the unique application of a robust assessment metric to permit identification of the features
