基于临近相似性考虑的犯罪热点密度图预测准确性比较——以DP半岛街头抢劫犯罪为例

在无时空考虑的密度估计算法基础上,分别加入了案件点之间的时间临近相似性、空间临近相似性和时空临近相似性的考虑,利用DP半岛2006~2007年的街头抢劫犯罪数据为基础计算无时空临近相似性、时间临近相似性、空间临近相似性和时空临近相似性4种不同算法所得到的犯罪热点图,并以之预测2008年的街头抢劫。通过Natural breaks（Jenks）分级方法和等比例面积选取两种方式来划定热点区域进行预测并进行PAI指数得分比较,结果表明时空临近相似性的密度估计算方法在犯罪预测的优势比较显著。

The Comparison of Predictive Accuracy of Crime Hotspot Density Maps with the Consideration of the Near Similarity:A Case Study of Robberies at DP Peninsula

Crime hotspot mapping can forecast the occurrence of potential offense and direct the police activities to restrain the offense by revealing the spatial pattern of the offense. Density estimation usually presents the most significant outcome of the potential offense among all the hotspot mapping methods, for the street crimes in particular. Compared with the density estimation algorithm without the spatial or temporal near similarity, the temporal near similarity, spatial near similarity and spatio-temporal near similarity among offense points were illustrated in this article. With street robbery data of DP Peninsula during the period from 2006 to 2007, four different crime hotspot maps based on non-spatial-temporal near similarity, spatial near similarity, temporal near similarity and spatial-temporal near similarity were produced respectively. Moreover, with the validation data of 2008 cases,two different classification methods, Natural breaks (Jenks) classification and Equal proportional selection by area were utilized to delimit the hot area and then the comparisons among the scores of the Prediction Accuracy Index (PAI) of those four different crime hotspot maps were implemented. The results presented that the density estimation method based on the spatial-temporal near similarity holds significant advantages when predicting the potential offences.