447 / 2019-02-27 21:35:57

Analysis for a Relationship between Risk of Burglary and Urban Space Composition: Case Study Focusing on 23 Wards and Urban Area of Tokyo

Crime mapping,Crime prevention,Burglary,Generalized linear mixed model,Urban space,Tokyo

全体主题 > Safety & Health Care Systems

In recent years, the number of crimes has been decreasing in Tokyo because of the installation of surveillance camera system and utilizing Geographic Information System(GIS) with web service to announce criminal information to the public etc., however decrease ratio of crimes also have been decreasing every year. Roughly 5000 of burglaries have occurred in the cities area and the 23 wards of Tokyo every year. On the other hand, National Police Agency interviewed experts involved in criminology, computer science and urban engineering so that they try to enhance crime prevention activities through the using of Information and Communication Technology (ICT). Consequently, in this research, focusing on 23 wards and the cities area of Tokyo, authors aim to identify the relationship between risk of burglary and urban space composition.
In this research, authors used the unit of small area: Cho-Cho-me, hereinafter, named CCMs, which is a familiar geographical unit for Japanese town, and it has been used in census. Authors examined the factors that have affected crimes using generalized linear model (GLM) and generalized linear mixed model (GLMM). Firstly, we collected the crime data from Tokyo Metropolitan Police Department`s web site which describe the total numbers of residential burglaries and store & office burglaries for the areas of Tokyo for 9 years, from 2009 to 2017. We used the data for an objective variable. Besides, we also used the data set which describe urban space composition from the Bureau of Urban Development of Tokyo Metropolitan Government, the location of public facilities and major road maps from the Ministry of Land, Infrastructure, Transport and Tourism, and major commerce & industry of CCMs from the Ministry of Economy, Trade and Industry. Secondly, we processed those data and calculated the explanatory variables, for example, mean Floor-Area Ratio of CCMs, distance to the nearest station, population density of CCMs, component ratio of commerce and industry at CCMs.
We had examined the statistical characteristics of distribution of burglaries so that we can analyze the relationship between risk of burglary and urban space composition by GLM and GLMM. As a result, we regarded that the both burglaries follow a negative binomial distribution or over-dispersion Poisson distribution. Then, we conducted the negative binomial regression with stepwise regression using the algorithm choosing the variables to minimize the Akaike's Information Criterion (AIC). Besides, we also conducted the over-dispersion Poisson regression with only the variables selected by the stepwise algorithm.
As a result of multiple regression analysis, we ascertained the model that the R-squared of over-dispersion Poisson model for residential burglaries equals 0.6201 and the model for store & office burglaries that the R-squared of over-dispersion Poisson model equals 0.6589. As a result, we identify some business and variables which describe about local population are influential in the occurrence of burglaries. Especially, the restaurants and other food-service businesses significantly increase the number of both burglaries while population density of CCMs decreases the number.
These results imply that the numbers of both burglaries are influenced not only by the shape of buildings or distance from public facilities but also by the local commerce and industry. Also, the model means that natural surveillance by local communities is one of the key factors that contributes to the enhancement of crime prevention performance for themselves.