时空犯罪热点预测模型研究

Research about Spatial-Temporal Forecasting of Crime Hotspots

统计分析表明,抢劫、抢夺、盗窃等街头犯罪具有时空聚集的特性,表现为时空热点,而以往的犯罪预测方法主要以统计理论为基础,侧重于对数据的描述,因此很难反映出犯罪时空热点变化的内在机制,因此也就无法给出较好的预测效果。日常活动理论指出,犯罪活动是由罪犯主体、受害主体和警察主体在时间和空间上出现交集而产生的,因此,通过模拟三种主体在一定时空条件下的运行就可得到犯罪活动的时空分布,进而得到犯罪的时空热点。基于日常活动理论和目前基于Agent的犯罪模拟研究,提出了一种具有信息反馈机制的时空犯罪热点预测模型,该模型在原理上能够反映犯罪的时空变化特性,进而预测出时空犯罪热点的变化。

The statistic analysis shows that the crime of robbery,snatching and burglary has the features of distributing by time and space,and then forms the hotspots.The traditional spatial-temporal forecasting of crime is a description predicting method based on the theory of statistic,so it is not possible to reflect the mechanism of hotspots＇ formation,and could not make forecasting well.The Routine Activity Theory points out that crime happens when three agents who are offenders,victims and police make convergence on space and time.Therefore,under some certain circumstance,the hotspots of crime could be modeled with the agents.A spatial and temporal forecasting model which has the mechanism of feedback was given out.This model can reflect the characters of crime variation by space and time in nature,and then it could make prediction of hotspots of crime.