Spatio-temporal Statistical Analysis And Influencing Factors Research On Property Cases In Xi’an

The property crime is a long-term objective social phenomenon,which seriously damages public safety and affects social stability.Research shows that crime does not happen randomly,instead,there is a certain regular pattern in time and space.Due to the confidentiality of crime data,there are few relevant empirical analyses.Therefore,it is necessary to conduct a more in-depth study of criminal activities in our country through empirical analysis,summarize its laws,and study the environmental factors that affect the spatial distribution of crimes,provide data support for crime prevention and control.Taking the police data from 2018.11.01 to 2019.11.26 in Xi’an as the research object,and the administrative streets(townships)as the research unit,statistical analysis of property cases was carried out in the two dimensions of time and space,and the correlation between street environment influencing factors and the number of street crimes was quantitatively explored.First of all,the time of crime is divided according to different scales,and the distribution characteristics of property cases on the scales of season,month,week,and hour are respectively explored.Secondly,based on GIS technology,the spatial global autocorrelation and spatial local autocorrelation methods are used to explore the spatial agglomeration patterns of property cases and identify the cold and hot areas of crime.Finally,27 factors of street environmental were selected as explanatory variables,and the number of street cases was used as the explained variables.Use multiple linear regression,stepwise regression,ridge regression,Lasso regression,random forest and XGBoost models to sort the importance of variables.Then compare the prediction results of each model to select the better performing model,and further analyze the explanatory variables with greater influence based on the ranking of the model.The study found:(1)In the time dimension,there is not much difference on the seasonal scale,and the number of cases in the season with higher average temperature is slightly higher than that in the season with lower average temperature.On monthly scale,the distribution of crime is relatively even,and the decline in February may be related to the Spring Festival.On weekly scale,the average number of cases during the week is higher than that at the weekend,with the highest number of incidents occurring on Thursday.During the day,it is generally characterized by being higher during the day than at night,with peaks at 8 o’clock and 18 o’clock,which belong to the peak period of commuting and conform to people’s daily behavior.(2)In the spatial dimension,there is obvious agglomeration,and it mainly presents a high-high aggregation situation.The high concentration area is the central area of Xi’an City,which mainly includes the Xincheng Diatrict,Yanta District,Lianhu District,Beilin District,and some streets in Chang’an District and Weiyang District.(3)The Lasso regression and XGBoost performed well in the prediction.The combination of the two models found that both street construction factors and street social factors have a certain impact on the number of cases.Among them,street construction factors with a greater degree of influence include:accommodation services,life services,transportation facilities,shopping services,scenic spots,land area,etc.The street social factors that have a greater impact include:the street’s 15-65 year-old population,the street’s female population,the number of local residents,and the total GDP.Among them,the land area and the number of local residents in the street are negatively correlated with the number of street cases,while the others are all positively correlated.