Predictive Modeling Combining Short and Long-Term Crime Risk Potential, Final Report
"This project developed a technology
capable of predicting future crime-risk potential based on a number of
theoretical approaches for understanding localized spatial crime
patterns.
The project had three goals. First, it
aimed to determine the way fundamental demographic correlates of crime
are linked to next year’s crime levels, even after controlling for this
year’s crime levels. Second, the study examined the role of near-repeat
crime events that are indicative of a short-term change in relative
risk of crime. Third, it developed a computer program that allows for
crime predictions based on the theoretical approaches presented. The
software is intended for use by cities and jurisdictions across the
United States. The project used crime data and Census information for
the City of Philadelphia, PA. For four crime types (robbery, aggravated
assault, burglary, and motor vehicle theft), a model that included
demographic structure and earlier crime from the previous year provided
by far the strongest combination of accuracy and parsimony. Lower
volume crime types (homicide and rape) were also predicted as well as,
or better than, the other four crime types in using the
demographics-only model. A model that combines community structural
characteristics, crime counts from the previous year, and an estimate of
near-repeat activity produced the best results overall. This indicates
that small-scale, short-term crime occurrences reflect a complex mix of
near-term crime continuities, ecological crime continuities, and
ecological structure. Work remains to be done in identifying the
processes that maintain these ecological crime continuities, as well as
the processes that generate the unfolding ecological discontinuities.
The authors note that the processes described ignore offender
characteristics, such as race, while focusing on locations of criminal
victimization."
View the Report
No comments:
Post a Comment