Analysis To Offenders' Journey-to-crime Pattern And Anchor Point Prediction

The journey to crime travel behavior(e.g.,travel distance and travel pattern)is typically associated with the specific features of the offenders,crime time,and crime environment.The majority of the papers on the journey to crime travel behavior do not use China's crime data and cannot provide practical implications in China.Our paper advances the literature on the journey to crime in three ways.First,we establish a modeling framework for the prescription and prediction of the offenders' travel behavior.Second,we uncover the effect of the offenders' personal features and the temporal and spatial features of crimes on the journey to crime travel behavior.Third,we adopt machine learning to predict the anchor point of offenders.Our result reveals the temporal and spatial features of crimes.The temporal feature of crimes can be characterized by the month,week,and hour of the crime.The main spatial feature of crimes is clustering.The cluster centers typically have high population density,high population mobility,and highly convenient traffic.By the time and space proximity analysis,we show that there are significant time and space proximity in the theft cases of electric bicycles.Besides,the offender 's features,such as gender,age,hometown,and crime experience,have a significant impact on the journey to crime distance because distinct personal features will lead to different perceptions of risks and costs.Specifically,male,middle-aged,foreign offenders with crime experience tend to take a longer journey to crime.Offenders with better physiological characteristics(e.g.,less knowledge of the environment)and more crime experience will have a larger travel space buffer,where offenders are less willing to commit to crime.Further,both the offenders' features and the temporal and spatial features of crimes have a significant impact on the journey to crime distance with varying magnitudes.Finally,we find that our prediction model based on China's crime data is more accurate than the traditional crime geographic profile prediction model.When applying machine learning to predict the type of the offenders' anchor point,we find that the decision tree algorithm achieves higher prediction accuracy and outperforms the other algorithms.Our work contributes to the growing empirical analysis of geographic profiling in China.The improved prediction model of the offenders' anchor point provides valuable investigation tools for the Public Security Department of China.