Developing a fractal model for spatial mapping of crime hotspots

Determining the precise area of urban crime hotspots, where the crime rate is significantly different from the surrounding area, is one of the essential applications of crime analysis. Besides, the classification of the spatial distribution of crimes can make spatial analysis more applicable to the police. In this regard, the present study examined the spatial distribution of five types of burglary in Zanjan city, NW Iran. Kernel density estimation (KDE) was initially applied to identify the crime distribution in the raster format. Because of producing a continuous-value surficial signature, the KDE model is not able to determine the exact borders of different crime density. To address this issue, the concentration-area (C-A) fractal model is proposed in this study as a frequency-spatial method for class designation of crime density. Thus, based on the interpolated maps derived from KDE, the C-A log-log plots consisting of the values of the gridded density maps of burglary crimes versus their occupied area were generated. Then, different threshold values and their corresponding criminal classes were delineated and mapped according to the drawn log-log plots. To evaluate the performance of the proposed fractal model, two GIS-based models, namely natural breaks classification (NBC) and equal interval classification (EIC), were generated over the KDE values. Then, a success rate curve was plotted as a validation method for quantitative evaluation of generated models based on the crime events. The obtained results revealed the benefit of the C-A fractal model in distinguishing different discretized classes of crime density over two other models. This can help to select the appropriate strategy to control and prevent crime occurrences in every crime density class.