The development of Urban Crime Simulator

Based on routine activities theory, deviant places theory, and neighborhood life cycle concepts, this paper describes the development of an Urban Crime Simulator (UCS) that was developed to allow estimation for changes in property crime rates in urban neighborhoods to be made when changes in the characteristics of the neighborhoods that are known or can be projected. The developed simulator is fully integrated with GIS-formatted data and operational in GIS environment. It enables users the flexibility of choosing neighborhood attributes that best fit their experience and knowledge of local neighborhoods. In addition, the selection of neighborhood attributes to be included in the simulation can be made based on particular criminological theories or expert experience that best fit localized trends.

Using the concepts of neighborhood life cycle, UCS first classifies urban neighborhoods into a number of clusters with minimized in-cluster variation and maximized between-cluster variation. When a chosen neighborhood is updated with projected changes in its attributes, UCS finds a neighborhood in the area that has the closest attribute profile with the changed profile of the neighborhood being simulated. The estimated changes in crime rates for the updated neighborhood are derived from statistical analysis of the neighborhoods in the cluster that the most similar neighborhood belongs to.

To assist users in better using UCS, a regression modeling tool is included to show the degree to which the variation in crime rates among neighborhoods is explained by the variations of the included neighborhood attributes. The Urban Crime Simulator is able to suggest an optimal number of clusters to be used in the simulation. In addition, the Urban Crime Simulator provides tools for calculating correlation between selected neighborhood attributes to avoid co-linearity. The simulator also includes tools for calculating global and localized spatial dependency of neighborhood attributes to assist users better understand the neighborhoods and their attributes. Essentially, users are given a set of tools to explore how their urban neighborhoods vary among themselves with selected attributes.

The developed urban crime simulator provides an efficient way to estimate possible changes in property crime based on known, projected or simulated changes in selected neighborhoods. With (1) a flexibility of users to select and include neighborhood attributes in the simulation, (2) a flexibility of accepting GIS data at different geographical scales, and (3) the tools assisting users in analyzing neighborhood attributes, the Urban Crime Simulator allows the users to fully incorporate appropriate/relevant criminological theories, their experience and expertise of local trends in the process.