犯罪预测效果指数构建与应用

Construction and Application of Crime Prediction Effectiveness Index

针对现有犯罪预测评价指标的不足,构建了更为合理的犯罪预测效果指数(Prediction effectiveness index,PEI).并通过对不同单元网格面积和预测热点数下的主要犯罪预测评价指标进行对比分析,验证了犯罪预测效果指数的合理性.在此基础上,提出了一种能提升犯罪预测效果的动态阈值犯罪预测方法,以WH市2015年1月至2019年6月入室盗窃犯罪数据为实证案例,使用长短期记忆模型进行测试验证.实验结果表明,相较其他犯罪预测评价指标,犯罪预测效果指数考虑的因素更全面,更合理.提出的动态阈值犯罪预测方法比传统固定阈值方法的实验效果平均提升了4.74%.

Aiming at the deficiency of the existing crime prediction evaluation index,a more reasonable Prediction Effectiveness Index(PEI)is constructed in this paper.The rationality of the crime prediction effectiveness index is verified by comparing and analyzing the main crime prediction evaluation index under different cell grid area and prediction hotspot number.On this basis,a dynamic threshold crime prediction method which can improve the effect of crime prediction is proposed.Taking the burglary crime data of WH City from January 2015to June 2019 as an empirical case,the Long Short-Term Memory model is used to test and verify.The experimental results show that compared with other crime prediction evaluation indexes,the crime prediction effectiveness index takes more comprehensive and reasonable factors into consideration.The dynamic threshold crime prediction method proposed in this paper is 4.74%higher than the traditional fixed threshold crime prediction method.