Identifying Crime Patterns Of Bus Pickpocketing Using Weighted Spatio-temporal Association Rules

In recent years, the number of bus pickpocketing is on the rise. Bus pickpocketing not only caused economic loss to the victims, but also endangered the physical safety of them, which caused great harm. In order to prevent and combat it, it needs to identify the crime patterns effectively. However bus pickpocketing may occurred when the bus is moving, it is difficult to define a location and crime time for this and traditional methods of crime analysis often focus on spatial or temporal properties of crime separately, so this paper introduced weighted spatio-temporal association rules mining to find out the spatio-temporal crime patterns of bus pickpocketing. It can be dug out through four steps.Firstly, temporal and spatial granularity division:dividing the main bus running time into equal unit of 2 hours named bus time then coded, and dividing the bus routes into bus sections by bus stops. Secondly, spatial analysis and time merge:extracting bus sections and crime time, and merging crime time with bus crime. Thirdly, as the crime rate of each bus section is different, the contribution to the result will also be different, this paper assigned each bus section a weight according to their crime rates. This paper assumed each bus section of a case has the same crime rate. Finally, weighted spatio-temporal association rules mining:using matrix based ICApriori algorithm and ICFP-growth algorithm to find out the spatio-temporal crime patterns of bus pickpocketing and visualizing the results on maps using GIS technology.Mining and map visualization results can be used to assist the police investigating and preventing bus crime, and optimizing the police deployment. The results prove this mining model has the following characteristics:1) It is creative to extract bus sections by bus stops.2) It is more realistic to assign a weight to each bus section according to the crime rate.3) The matrix based ICFP-growth algorithm reduced the number of database scanning and is more efficient than ICApriori algorithm.4) Users can observe and analyze the results more intuitive and get more useful information through map visualization.