Visualization of Crime Trajectories with Self-Organizing Maps: A Case Study on Evaluating the Impact of Hurricanes on Spatio- Temporal Crime Hotspots

The exploration of the spatial relationships between crime incidents, the socioeconomic characteristics of neighborhoods, as well as physical and structural compositions of the urban landscape is an ongoing research issue in Geographic Information (GI) Science. Spatial data mining tools improve the ability to gain knowledge from geographic data and help to understand spatio-temporal processes that contribute to the presence or absence of criminal offenses. However, most of the currently available tools focus either on the spatial, the temporal, or a combination of both aspects. But crime has a spatial and temporal component in a multidimensional attribute space. Therefore, it is reasonable to combine all these aspects within one analytical framework. This paper presents such a methodology to explore crime patterns and their spatial and temporal behavior within their socio-economic and environmental neighborhoods. The framework consists of three complementary techniques: A spatio-temporal scan statistic to detect crime hotspots, a growing selforganizing map (SOM) to analyze attribute properties of the neighborhoods, and a mapping of crime hotspot trajectories onto different SOM visualizations. The case study uses burglary locations from Houston, Texas, from August to October 2005.