Characterizing Crimes from Web
Crime prevention requires the effective use of police resources, which demands the access of criminal information for planning security actions. The number of crime occurrences is higher than the official reported numbers. Many victims do not report crimes directly to the security agencies. Instead, they prefer to anonymously report using different channels, such as the Web. In this article, we introduce our approach to characterize crimes reported in the Web. Particularly, we collect criminal data from popular websites that store crime occurrences, and we use clustering analysis to discover crime patterns on the collected data. Applying our approach to a popular Brazilian crime report website, we observe that more than 41% of the crimes were not reported to the security agencies, and most of them are thefts and robberies occurring at night and dawn. In addition, minor offenses present different patterns of serious crimes. Moreover, crime patterns are different in rich and poor neighborhood.