A large body of research has found that crime is much more likely to occur at certain places relative to others. Attempting to capitalize on this finding to maximize crime prevention, many police administrators have sought to narrow their department’s operational focus and allocate resources and attention to the most problematic locations. However, in the face of a growing number of technological advances in crime forecasting that have facilitated this trend, it is still unclear how to best identify the most appropriate set of places to which resources and attention should be directed. Our goal was to examine this issue by exploring the ways in which spatial vulnerabilities and exposures could be used to identify the best target areas for policing. Using the Theory of Risky Places as a guide, we employed kernel density estimation (KDE) to measure crime exposures and risk terrain modeling (RTM) to identify crime vulnerabilities with the expectation that crime would be predicted more accurately by integrating the outputs from these two methods. To test this hypothesis, our analysis utilized 1 year of data on street robbery in Brooklyn, New York. A common metric, the prediction accuracy index (PAI), was computed for KDE, RTM, and the integrated approach, over 1 month and 3 month intervals. We found that the integrated approach, on average and most frequently, produces the most accurate predictions. These results demonstrate that place-based policing and related policies can be improved via actionable intelligence produced from multiple crime analysis tools that are designed to measure unique aspects of the spatial dynamics of crime.

中文翻译：

---

利用漏洞和暴露来改善抢劫预测和目标区域选择

大量研究发现，相对于其他地方，在某些地方犯罪的可能性更大。为了利用这一发现来最大程度地预防犯罪，许多警察管理人员试图缩小部门的工作重点，并将资源和关注点分配到最有问题的地点。但是，面对犯罪预测方面越来越多的技术进步，这种趋势已成为现实，但仍不清楚如何最好地确定最合适的资源和关注场所。我们的目标是通过探索空间漏洞和暴露可用于确定最佳警务目标区域的方式来研究此问题。以危险场所理论为指导，我们使用核密度估计（KDE）来衡量犯罪暴露，并使用风险地形模型（RTM）来识别犯罪脆弱性，并期望通过将这两种方法的输出相结合，可以更准确地预测犯罪。为了检验这一假设，我们的分析利用了纽约布鲁克林街头抢劫案的1年数据。在1个月和3个月的时间间隔内，针对KDE，RTM和集成方法计算了一个通用指标，即预测准确性指数（PAI）。我们发现，综合方法平均且最频繁地产生最准确的预测。这些结果表明，通过多种犯罪分析工具产生的可操作情报，可以改善基于地点的警务和相关政策，这些工具旨在测量犯罪空间动态的独特方面。