Integration and Evaluation of Social Media Data into Crime Prediction Models

Within the overall framework of Geospatial Technology and GIScience in Spatial Crime Analysis and Mapping I am proposing an innovative research topic that combines information from social media data with crime prediction models.  Specifically, information included in social media data that is of relevance to the police (case-related information, such as locations, social networks, etc.) to solve crimes is extracted and integrated into crime prediction models.  The overall goal is to evaluate the impact that social media information has on the accuracy of forecasting crime events.

Spatial and/or temporal crime prediction models have become very popular over the past five to ten years, yet very little is known about which models work best.  For this reason, I am proposing first, a comprehensive testing and evaluation of crime prediction models and software, second, the recommendation of a suite of spatial and/or temporal models to be applicable for different crime types and in different cities and countries (mostly Austria), and third, the potential impact that the integration of social media information into crime prediction models has on the quality of the predictions.

The regional focus for this research topic will be Austria.  The main reason is that many of the theories, concepts, and guidelines for crime prediction have been exclusively developed in the US and in the UK, using mostly US and UK crime data.  Consequently, such theories, concepts, and guidelines may not be directly applicable to Austria or the German-speaking countries of Europe.  For this reason, there is a need to revisit and possibly adapt existing theories, concepts, and guidelines for crime prediction with comprehensive, systematic, and scientific studies using crime data from Austria and other German-speaking countries.  A second reason for choosing Austria as the study area is that crime data reported to the police have been collected for all of Austria at the address-level since 2004.  These crime data will be available for this research and allow a comprehensive and systematic study of spatial and/or temporal forecasting models.

Possible research questions include, but are not limited to:

·         What is the “best” spatial and/or temporal crime prediction model, in general and for the different crime types, in particular?

·         What impact, if any, has the implementation of social media information into models that predict the spatial and/or temporal occurrence of crime events?