Research On Spatio-temporal Patterns Of Burglaries Based On Hierarchical Bayesian Models

Crime is an existing phenomenon in the process of social development.The forms and approaches of committing a crime are complicated and changing with the development of social economy.In addition,the number of criminal cases is constantly growing up.Crime has seriously affected the social stability,arousing great concern from the government,scholars and the public.As criminal activities always occur in a certain space,crime geography,focusing on the spatial aspect of crime,has become a crucial part of crime studies and is playing an increasingly significant role in crime prevention and control.The rapid development of GIS theory and technology has deeply promoted crime geography studies.However,crime geography research in China is still on its infancy compared to that abroad.The contents and the depth of the research need to be enhanced further.Meanwhile,the research methods adopted should also be improved and innovated.Burglary is the most common crime worldwide.It occurs frequently and the consequences are usually serious.In this context,this paper first summaries the present situation of crime geography studies.After that,based on theories related to crime geography,it utilizes GIS,spatial statistics and spatio-temporal modeling approaches to deeply analyze the spatio-temporal patterns of burglaries in Jianghan District,Wuhan,China.Results of the paper can provide decision support for police management,police resource allocation and crime prevention and control.More specifically,the main contents of the paper are as follows:(1)Analyzing the spatio-temporal distribution characteristics of burglariesFirst,the paper utilizes exploratory data analysis methods to analyze the features of the monthly,weekly and daily changes of burglaries in Jianghan District.Second,the paper utilizes exploratory spatial data analysis methods to analyze the overall distribution characteristics of burglaries and the temporal variations of standardized burglary ratio in each neighborhood.Spatial autocorrelation is also explored.Third,this paper introduces a hierarchical Bayesian spatio-temporal modeling approach to analyze the patterns of burglary risks over time.The Bayesian spatio-temporal model overcomes the deficiencies of standardized burglary ratio and can tackle spatial autocorrelation.By examining the posterior probability of differential trend being greater than zero for each neighborhood,the paper explores the strength that burglary risk in each neighborhood having an upward trend compared to the risk variation for the whole region.(2)Exploring the influence of neighborhood environmental characteristics and backgroud effects on burglariesBased on theories related to crime geography,we identify some environmental variables at the neighborhood level as potential risk factors from the available data.Meanwhile,in order to overcome the shortages of non-spatial models,the spatial lag models,the spatial error models,the geographically weighted regression models and other frequency-based models,we adopt Bayesian models.Several Bayesian models are utilized,analyzed and compared to explore the way of dealing with overdispersion,spatial autocorrelation and the small number problem.In addition,we propose three models that account for background effects to investigate whether they can significantly improve the original model.According to the results of the models,we choose the best model as the final model by comparing and assessing all the models.After that,the hot and non-hot spots of burglary risks are identified by the best model.Results show that two of the three models accounting for background effects have the same goodness of fit as that of the original model.Furthermore,models accounting for background effects can be extended to conduct multi-scale joint analysis from a spatial perspective.(3)Studying the classification of neighborhoods based their burglary risksThis paper proposes a classification rule for the neighborhoods in the study region.The rule considerers not only the present situation of the burglary risk in each neighborhood,but also considers its future variation trend.Neighborhoods are classified into six types.One can easily judge the basic situation of burglary risk in any neighborhood by the type the neighborhood belonging to.(4)Joint modeling of multiple crimesThe existing criminological research typically focuses on one type of crime.This paper proposes a method that jointly models the counts of multiple types of crime.The method pools all the available data from different crime sources and accounts for the correlation between different types of crime.Analysis conducted by joint modeling of multiple crimes can better reveal the true clustering of crime risks than that conducted by modeling each crime type separately.