| Resumen |
In this paper, a novel similarity function is used to identify hot-spots of criminal activity in large crime-datasets. This function considers the space and times when each crime was committed, as well as some elements of the perceived modus operandi of the perpetrator, in order to compare specific crime patterns and then cluster them using a density-based clustering algorithm. The clusters so formed are then graphically shown to the crime analyst using diverse GIS-tools, in order to provide him/her with high quality information about the current state of criminal activity. Several experiments performed, as well as a case-based comparison with previously published similar proposals, yield significant advantages of the proposed function over classical Euclidean-distance comparisons and other space-time similarity functions. |
