The authors of the article Feature engineering for crime hotspot detection state that cities with a complex infrastructure require ‘smart’ systems of crime detection; they claim the hotspot detection approach to be highly effective. The reviewed article focuses on two cities: San Fransisco (US) and Natal (Brazil); the authors analyzed “spatio-temporal and urban features” of both cities in order to evaluate their significance in crime detection (Borges et al., 2019, p. 2). Such factors as the number of odds per area, prosecution rate per area, road types, police types, and time features have been taken into account for a proper depiction of the criminal hotspots (Borges et al., 2019). The interlink between particular infrastructural specificities and areas with high crime records has been analyzed (Borges et al., 2019). In addition, “the urban space was discretized into cells using the k-means clustering algorithm”; after that, the cells were “labeled into criminal hotspots or inconspicuous cells” (Borges et al., 2019, p. 4). As a result, this approach allowed the researchers to create a map that uses various urban features to effectively detect criminal hotspots in relation to both cities’ infrastructure.
The main findings of the reviewed study have shown that there is a clear interlink between particular urban features of the city and crime levels. Such specificities as the capacity or length of a road or pedestrian area play an essential role in forming local crime hotspots (Borges et al., 2019). In addition, the features of a particular city were also considered. Thus, the “presence of college in the area” had a correlation with local crime levels in Natal (Borges et al., 2019, p. 8). Thus, the method of city infrastructure analysis, together with the hotspot detection approach, has proven to be quite effective.
References
Borges, J., Ziehr, D., Beigl, M., Cacho, N., Martins, A., Sudrich, S., Abt, S., Frey, P., Knapp, T., Etter, M., & Popp, J. (2017). Feature engineering for crime hotspot detection. 2017 IEEE SmartWorld, 1–8. Web.