›› 2016, Vol. 28 ›› Issue (8): 125-132.

• 应急管理专辑 • 上一篇    下一篇

基于和声搜索算法优化支持向量机的突发暴恐事件分级研究

王雷1,2, 王欣1,3, 赵秋红1   

  1. 1. 北京航空航天大学经济管理学院, 北京 100191;
    2. 中国刑事警察学院治安学系, 沈阳 110035;
    3. 中国刑事警察学院公安情报学系, 沈阳 110035
  • 收稿日期:2015-07-08 出版日期:2016-08-28 发布日期:2016-09-24
  • 通讯作者: 赵秋红(通讯作者),北京航空航天大学经济管理学院教授,博士生导师
  • 作者简介:王雷,北京航空航天大学经济管理学院博士后,中国刑事警察学院治安学系讲师,博士;王欣,北京航空航天大学经济管理学院博士后,中国刑事警察学院公安情报学系讲师,博士
  • 基金资助:

    国家自然科学基金项目(91224007);辽宁社会科学规划基金项目(L15AGL016);公安部公安理论及软科学研究计划(2016LLYJXJXY032);辽宁省大学生创新训练计划项目(201610175000030)。

Optimizing Parameters of Support Vector Machine Using Harmony Search Algorithm for Emergency Classification of Terrorist Attacks

Wang Lei1,2, Wang Xin1,3, Zhao Qiuhong1   

  1. 1. School of Economics and Management, Beijing University of Aeronautics and Astronautics, Beijing 100191;
    2. Department of Public Order, National Police University of China, Shenyang 110035;
    3. Department of Public Security Intelligence Science, National Police University of China, Shenyang 110035
  • Received:2015-07-08 Online:2016-08-28 Published:2016-09-24

摘要:

突发暴恐事件分级具有重要作用,能够保证预案合理执行和应急资源优化配置。提出基于和声搜索算法优化的支持向量机的分级模型,用于突发暴恐事件分级研究。和声搜索算法优化支持向量机参数,支持向量机提供学习和曲线拟合,同时根据准确度、精确度和敏感度指标评估混战智能分类模型的绩效。利用全球反恐数据库中2008年至2013年我国暴恐事件数据进行测试,并与支持向量机、分类与回归决策树(CART)和C5.0方法进行对比,结果表明分级方法可行且有效,能够为突发暴恐事件管理提供预警和决策支持信息。

关键词: 突发暴恐事件, 分级, 和声搜索, 支持向量机

Abstract:

The classification on terrorist attacks plays an important role to ensure optimal allocation of emergency resources and reasonable implementation for emergency plan. This study proposes a model to integrate a harmony search with a support vector machine (SVM) to research classification of terrorist attacks. The support vector machine provides learning and curve fitting while harmony search optimizes support vector machine parameters. Measures in term of accuracy, precision and sensitivity are used for performance evaluation of proposed hybrid intelligence classification model. The data of global terrorism database from 2008 to 2013 in china is used for testing, and experimental comparisons indicate that HS-based SVM achieves better accuracy compared to SVM, CART and C5.0. Experimental results show that HSSVM is a feasible approach dealing with emergency classification problems, and provides warning and decision support information needed to manage emergency terrorist attacks.

Key words: terrorist attacks, emergency classification, harmony search, support vector machine