Research On Crime Hot Spot Prediction Of City Stealing Cases With Spatial And Temporal Nonstationary

The analysis of spatial and temporal nonstationary could not only reveal the spatial and temporal distribution of geographical factors,but also explain the dynamic process of social phenomena and the environment.Geographically and temporally weighted regression(GTWR),based on the geo-weighted regression model,takes the spatial and temporal impact into consideration,integrates the geographical and temporal effects into the model and solves both spatial and temporal non-stationarity issues.The distribution of crimes in the geographical and temporal is not uniform or random,but presents regularity,which is specifically embodied in temporal periodicity and spatial clustering.The analysis of criminal hotspots can effectively express the temporal and spatial clustering of crimes,visually display the space-time position(time and place)of crimes,and predict the crime hotspots,which is of guiding significance for public security.Based on regression modeling method,this paper establishes a relational model between environmental impact factor and crime hotspot,and uses environmental impact factors to predict spatio-temporal hot spots.Geographically and temporally weighted regression model,which uses regression modeling method,incorporates time and space factors into the regression model to solve the spatiotemporal nonstationary in the hot spot of the crime.The research work and content are mainly the following:(1)Space-time Nonstationary Analysis of Theft CasesExploratory analysis was conducted on the point data of theft cases from the time and space dimensions,and this paper analysis the spatio-temporal distribution of the theft cases to find the potential law of the theft point data.The nonstationary of the theft cases in space and time is tested through the significance analysis,and this paper consider whether to use the Geographically and temporally weighted regression model,which takes account of the non-stationary time and space when using the regression model to predict.(2)Prediction of crime hot spots by using spatio-temporal weighted regression modelThis paper proposes the prediction method of the crime hot spot based on spatiotemporal weighted regression model,expands the space kernel function as a spatiotemporal kernel function,decides the optimal space bandwidth and optimal space-time factor by using CV method,and finally the spatio-temporal distribution of theft cases and the spatio-temporal weighted regression model of the crime influence factors are constructed.The spatio-temporal geographical weighted regression model is used to predict the time and space distribution of theft cases,and the kernel density function is used to identify the hot spots of the crime of theft,thus this paper realizes the prediction of the hot spot of the crime by using Geographically and temporally weighted regression model.(3)Evaluation of the prediction results of crime hot spots in theft casesThis paper analysis the spatio-temporal geographical weighted regression model's fitting results by using the variance analysis and goodness-of-fit analysis,and compares the advantages and disadvantages of multiple linear regression and geo weighted regression in the fitting of theft cases.The prediction accuracy and the prediction result of spatio-temporal geographically weighted regression model are analyzed.This paper uses spatio-temporal geographic weighted regression model to predict the crime hotspots by taking the theft cases in the core area of Suzhou city as an example.The fitting index of the predicted results was 0.4511,the prediction accuracy of the model is 9.08%higher than that of the geographically weighted regression model,the prediction result of model is better.