Study On The Regional Characteristics And Crime Behavior Patterns Of Criminals

In recent years,a lot of studies have focused on the relationship between urban crime and floating population,but most of them are from the perspective of demography and sociology to analyze and explain the reasons or causes of crimes committed by floating population,and lack of analysis on the regional characteristics of the domiciles and urban crime behavior of nonlocal criminals.It is helpful to expand the intervention of crime prevention from “place of crime”to “domiciles”,formulate more targeted positive guidance strategies such as publicity and education,and improve the accuracy and effectiveness of crime prevention.And it is helpful to find out the potential association of criminals and cases to reduce the scope of investigation of police in the process of solving cases by analyzing the characteristics of crime behavior of different criminals and establishing the relationship between them and the domiciles of criminals.At present,with the implementation of the strategy of “Intelligence-Led Policing”,intelligence research and analysis based on multi-source data plays an increasingly important role in public security work,and also provides important opportunities to explore the regional characteristics of criminals,crime behavior patterns and the relationship between the two.Taking the burglary cases in Beijing from 2005 to 2014 as an example,this paper analyzes the regional characteristics,spatial flow mechanism and crime behavior patterns of non-local criminals.On this basis,it studies the machine learning prediction model to determine the source identity of criminals,and develops the corresponding crime analysis software.The main conclusions are as follows:(1)In terms of regional characteristics of sources,Moran's statistics and other spatial analysis methods are used to find that there is a significant spatial agglomeration effect in the distribution of domiciles of non-local criminals in Beijing.The provinces and cities close to Beijing have always been the main sources of non-local criminals in Beijing from 2005 to 2014.However,the number of non-local criminals in the provinces and cities far away from Beijing is gradually increasing,showing an obvious trend of regional gangs.At the same time,there is a strong positive correlation between non-local criminals and floating population in Beijing.The mobility and aggregation of floating population increase the risk of urban crime and the difficulty of population management and control.(2)In terms of spatial mobility mechanism,based on multiple linear regression model and other methods,this paper analyzes the spatial mobility mechanism of criminals from the perspective of floating population management and service.Under the control of other conditions,the lower the per capita income level,the higher the education level,the lower thepopulation density,and the closer the distance to Beijing,the greater the number of floating population in Beijing from the overall average effect;from the perspective of local spatial effect,the social and economic conditions of different regions have spatial heterogeneity on the number of floating population in Beijing,which indicates that the spatial mobility mechanism of floating population from different regions has spatial differences.(3)In terms of crime behavior pattern,through the non-parametric test and other methods,it is found that the crime areas of non-local criminals spread from the main urban areas to the surrounding urban areas in Beijing,and the violence and gang trend of crime means are gradually increasing.However,with the development of time,the crime behavior patterns of the two criminal groups show a certain convergence.(4)In terms of the prediction model based on machine learning,the empirical results of different years' data show that taking the time,location type and means of crime as the input,using naive bayes,logistic regression,decision tree,random forest,the regional identity of criminals(local / non-local)can be identified with the accuracy of above 82%.Among them,the prediction accuracy of the model based on random forest is always the highest,which keeps above 85%.On the whole,based on the data of public security departments,this paper deeply explores the regional characteristics,spatial flow mechanism and crime behavior pattern of criminals,and studies the machine learning model applied to determine the source identity of criminals,which not only helps to understand the spatial flow and crime selection process of criminals for the public security departments,but also helps to expand the vision of crime prevention and reduce the scope of investigation,which has important theoretical value and practical significance.