Embedding and operationalising predictive analytics in policing

This thesis examines the development and operational use of predictive analytics in policing. The extent to which predictive analytics has become embedded in policing is of fundamental importance to questions about the more specific use of predictive analytics in a policing context. Before assessments and recommendations of proactive policing, crime reduction, or victim satisfaction can be made; police forces need to have appropriately and effectively embedded the use of predictive analytic model outputs within the working practices of their frontline users. From the literature reviewed, predictive analytics has become a buzzword in policing. However, there seems to be no clear standard of method of evaluating either the overall or individual use of predictive analytics in a policing context. Using Avon and Somerset Constabulary as a case study, this thesis has investigated the use of three unique predictive analytics models developed and used by the Constabulary. This thesis explored the use of the predictive model outputs, assessments of the predictive model, user implementation of model outputs, and the attitudes of staff using the predictive model outputs. This was achieved using a pragmatic research methodology, ethnography, field observation, qualitative semi-structured interviews, and an existing police user survey. This research has found that while predictive analytics to have potential for use in policing, more work needs to be done to formally embed predictive analytics into operational policing in Avon and Somerset Constabulary. Recommendations for police forces intending to introduce and use predictive analytics in policing include, creating a supportive infrastructure, and the need for education and training on model use and outputs as well as subsequent evaluation of outcomes and input back into the models