This document presents study results on predictive methods that enable law enforcement agencies to proactively address criminal activity.
Research objectives included devising predictive models that specifically addressed the needs of law enforcement; evaluating the effectiveness of the prediction models; and disseminating the most promising models for use by law enforcement agencies. This research is organized into three papers. The first paper provides the theoretical foundation for the new approach to crime event prediction, built out of results in space-time point processes. The fundamentals are reviewed and theoretical details are given. The second paper provides applications of the approach to a crime prediction problem in Richmond, Virginia. Breaking and entering events were examined and data were used from 1 week to predict both the next week and the next 2 weeks. Predictions were compared to those provided by several density estimation approaches. The use of feature data out-performed these estimates, suggesting the use of these data can improve the prediction of criminal events. The third paper provides an extension of the model to handle temporal features. The problem of measuring similarity between temporal features was discussed. It is shown how temporal features can be used with point process model to provide for both space-time attributes. The approach is tested using data from Richmond, Virginia. In some cases, the temporal features improved performance, but this was not always the case. A filter that identifies incidents that have low variance in certain temporal features is needed. This paper also provides information on the implementation of the prediction methodology.
Downloads
Similar Publications
- Importance of culture in measuring tribal crime seriousness: scoping review of crime seriousness indices
- Mitigating the Harm of Public Mass Shooting Incidents through Situational Crime Prevention
- Youth Bias-Based Victimization: Comparing Online Only, In-Person Only, and Mixed Online/In-Person Incidents