基于GRA-SVR的恐怖风险情报预测模型——以“一带一路”为例

Terrorism Risk Prediction Model Based on GRA-SVRTa——king "the Belt and Road" as an Example

[目的/意义]为实现对区域恐怖风险指数的短期预测,提出一种基于灰色关联(Grey relational analysis, GRA)支持向量回归(support vector regression, SVR)的恐怖风险预测模型。[方法/过程]引入科学知识图谱对恐怖风险要素进行识别,并根据风险要素确定衡量指标;利用灰色关联分析计算各影响指标与恐怖风险的关联度,筛选出主要的影响指标构建评估指标体系;通过网格搜索交叉验证寻优支持向量回归机参数,使用优化后的参数建立支持向量回归模型。[结果/结论]来自"一带一路"沿线国家2008-2018年数据仿真结果表明,基于GRA-SVR的恐怖风险预测模型具备更优的预测能力,该模型可在一定程度上弥补现有方法通用性不强、检测结果与恐怖风险实际发生的相关性不高等不足,为我国国家安全战略的制定和实施提供参考。

[Purpose/Significance]To realize the short-term prediction of regional terrorism risk index, this paper proposes a terrorism risk prediction model based on grey relational support vector regression.[Method/Process]Firstly,mapping knowledge domain is introduced to identify the terrorism risk elements,and the quantitative indicators are determined according to the risk elements.Then,the grey relational analysis is used to calculate the relation degree between each influence indicator and the risk of terrorism,extract the main indicators,and construct the assessment indicator system.In addition, the parameters of SVR are optimized by combining grid search with cross validation method,and the support vector regression model is established with the optimized parameters. Simulation is carried out using 2008-2018 years data from "the Belt and Road". [Result/Conclusion]Case analysis and comparison results show that GRASVR-based terrorism risk prediction model has good prediction ability.To a certain extent,this model can make up for the shortcomings of the existing methods,such as weak universality,low correlation between the detection results and the actual occurrence of terrorism.