Crime Mapping through Geo-Spatial Social Media Activity

The presence of crime is one of the major challenges for societies all over the World, especially in metropolitan areas. As indicated by prior research, Information Systems can contribute greatly to cope with the complex factors that influence the emergence and location of delinquencies. In this work, we combine commonly used approaches of static environmental characteristics with Social Media. We expect that blending in such dynamic information of public behavior is a valuable addition to explain and predict criminal activity. Consequently, we employ Zero-Inflated Poisson Regressions and Geographically Weighted Regressions to examine how suitable Social Media data actually is for this purpose. Our results unveil geographic variation of explanatory power throughout a metropolitan area. Furthermore, we find that Social Media works exceptionally well for description of certain crime types and thus is also likely to enhance the accuracy of delinquency prediction.