Determination of spatial factors affecting burglary crimes with CBS-supported machine learning methods and developing a crime forecast model

Crime events occur depending on spatial and non-spatial factors. In order to fight crime effectively, it is necessary to conduct crime analysis and examine the factors that cause crime. It has been observed that the number of studies examining the distribution of crime based on spatial factors is low in our country. With the integration of spatial analysis methods and machine learning of GIS, mathematical relations can be established between crime and crime-causing events. In order to deal with the hypothesis researched in this thesis, Kilis city center was determined as the study area and spatial data were obtained. Burglary crime events that took place in the years 2018-2019-2020 were used in the study. The density map of the theft crime in the city was produced by Kernel Density Analysis method. Then, the relationship between crime density and environmental and urban design criteria affecting crime was examined with machine learning and GIS-based analyzes. Input data (26 factors) for machine learning were obtained by arranging the spatial distribution of theft crimes and the spatial factors affecting the crime with GIS. For the machine learning model, the crime prediction model was created by analyzing with Random Forest, Gradient Boosting and Support Vector Regression algorithms. As a result of the analyses, the order of importance of spatial factors in crime formation and their effect coefficients on crime formation were obtained, and burglary crime risk potential maps based on spatial factors were created with the values obtained from the crime prediction model. The estimation model and maps obtained will contribute to the literature by presenting a dynamic approach with the advantage of technology in the land management process by showing the effect of the spatial setup on the crime potential.