Macro-Level Collision and Crime Analysis: Case Study for the City of Regina
Traffic collisions and crimes are issues of concern in most neighbourhoods and cities and they are certainly a concern for the City of Regina. The traditional approach to either preventing or reducing the severity of collisions and crimes has been a reactive one: identifying locations as problematic based on historical data before taking action. An advanced and recently introduced approach for dealing with the issues of collision and crime is the Data-Driven Approaches to Crime and Traffic Safety (DDACTS). DDACTS is a proactive, place-based approach that identifies problematic locations that require interventions. Results from a macro-level analysis are used for planning purposes. Traffic Analysis Zones for the City of Regina were considered in this research. Traffic Analysis Zones are a spatial aggregation of census blocks and are, in part, a function of population, used by city planners, for planning new neighbourhoods and resource allocation, as well as by transportation officials for tabulating traffic-related data. Traffic Analysis Zones level collision and crime prediction models have been developed to estimate safety and security effects of neighbourhood level land use, socio-economic factors, road network characteristics, and demographic variables on collisions and crimes. Furthermore, the Empirical Bayes technique are adopted to estimate expected frequencies of collisions and crimes. The expected frequencies are used in determining hotspots that require enforcement and countermeasures. The Negative Binomial modeling technique was adopted in this study to predict numbers of collisions and crimes. Models were calibrated and validated using multiple goodness-of-fit tests. Results from the goodness-of-fit tests were used as basis to determine the best model for predicting each type of collision and crime. Maps were then created to display both spatial patterns and spatio-temporal trends of collisions and crimes. Traffic Analysis Zones with significant frequencies of collisions and crimes, both separately and in unison, were then identified. Some of the conclusions drawn from the collision prediction models include: both intersection density and intersection road density had positive associations with collisions; and when comparing 3-leg and 4-leg intersections, 3-leg intersections had fewer safety concerns. Also, low density residential areas have collision reduction effects. Results from collision prediction models developed in this study can help transportation engineering officials, and city planners in traffic safety decision. At the planning stage of new neighbourhoods, the safety effects of individual predictors or sets of predictors can be determined by creating multiple scenarios that involve interested sets of variables. The developed crime models provided information about how land use type, socio-demographics, and residential land use type influence different crime types. Some conclusion drawn include the following: commercial areas and retail spaces were target areas for high numbers of violent crimes; high population density neighbourhoods attracted high numbers of crimes; higher numbers of residents within the age groups of 18 to 24 and 25 to 44 were positively associated with both violent and non-violent crimes; residents within the age groups of 44 to 65 as well as 65 years and over had a crime reduction effect, regardless of the crime occurrence type. Also, low density residential areas attracted many non-violent crimes; industry and office areas also attracted many non-violent crimes; and multiple or mixed land use areas also attracted a high volume of auto-involving theft crimes. The results of this research is intended to improve the lives of the residents of the City of Regina by providing tools that can be used to reduce traffic collisions and crimes.
Collision Crimes Prediction Analysis Empirical Bayes Data-Driven
Master of Science (M.Sc.)
Civil and Geological Engineering