Research On Spatio-temporal Distribution Characteristics Of Theft Crimes

With the development of our society, people’s life has earth-shaking changes and all kinds of social contradictions are constantly emerge. Crime is a heavy topic in the development of society, the crime may do harm to people’s property security and social order, and the number of crime is rising year by year. Scholars have found that the distribution of crime in time dimensions or space dimensions are not completely random, it has a certain distribution characteristics. Using appropriate methods can find spatio-temporal distribution characteristics of crime, it can provide important information for the public security personnel, It is not only related to business assessment and police resources deployment, but also affect the formulation and implementation of public security prevention and control strategy. Based on previous research status and key technology, this paper applying spatial autocorrelation, self-organizing maps and time series analysis to study the spatio-temporal distribution characteristics of crimes in Fuzhou city. The results can provide decision support to security personnel in handling the cases and reasonably distribute police resources. The main research contents and features as follows:(1) Summarized the technical route of research on spatial-temporal distribution characteristics. It contains clustering analysis and visualization based on statistical data of surface pattern, clustering analysis and crime hotspots recognition based on point pattern, and the time series prediction analysis of crime.(2) Research on spatio-temporal distribution and visualization based on statistical data of surface pattern. Analyzing spatial autocorrelation of surface pattern which based on the police station in main districts of Fuzhou, first using the global spatial autocorrelation to judge whether the cases are gathered, if it is gathered, then analyzing the distribution of gathers by using local spatial autocorrelation. Then using self-organizing map and U matrix method to analyze spatio-temporal multidimensional visualization based on the different types of theft within the police station, the results of the analysis can help the public security personnel to understand the distribution characteristics of crime which implied in the complex multidimensional data.(3) Hierarchical cluster and kernel density estimation are analyzed to study the distribution of point pattern. In main districts of Fuzhou city as an example. Hierarchical clustering can express the distribution of crimes by using the classification, Kernel density estimation can express continuous changes and accurate cluster center of crimes. Then comparing the results of the analysis. The results of the study clearly shows the spatio-temporal distribution of the research area.(4) The time series prediction analysis of crime. Using autoregressive integrated moving average model (ARIMA) to predict the future number of crimes. And the outcome of prediction is correspond to the actual value.