地球信息科学学报 ›› 2021, Vol. 23 ›› Issue (4): 584-592.doi: 10.12082/dqxxkx.2021.200199

• 地球信息科学理论与方法 • 上一篇    下一篇

顾及时空邻近的恐怖团伙关系发现方法

刘慧敏(), 刘青豪*(), 陈袁芳, 石岩, 邓敏   

  1. 中南大学地球科学与信息物理学院,长沙 410083
  • 收稿日期:2020-04-25 修回日期:2020-07-11 出版日期:2021-04-25 发布日期:2021-06-25
  • 通讯作者: 刘青豪
  • 作者简介:刘慧敏(1977— ),女,湖南宁乡人,博士,副教授,主要从事时空大数据融合与信息服务研究。E-mail: lhmgis@163.com
  • 基金资助:
    国家自然科学基金项目(41771492);湖南省自然科学基金青年基金项目(2018JJ3639)

A Method of Finding the Relationship between Terrorist Groups Considering Spatiotemporal Proximity

LIU Huimin(), LIU Qinghao*(), CHEN Yuanfang, SHI Yan, DENG Min   

  1. College of Geosciences and Information Physics, Central South University, Changsha 410083, China
  • Received:2020-04-25 Revised:2020-07-11 Online:2021-04-25 Published:2021-06-25
  • Contact: LIU Qinghao
  • Supported by:
    National Natural Science Foundation of China(41771492);Youth Fund Project of Natural Science Foundation of Hunan Province(2018JJ3639)

摘要:

现有恐怖团伙关系挖掘方法因其难以顾及恐怖组织特有的结构复杂、行动灵活、生存期长等特点而难以应用于反恐研究中。为此,本文从时空角度出发,通过引入时空邻近与团伙时空同现模式等概念,将多元独立分布的团伙关系发现问题建模为时空同现模式的频繁度评价问题,提出一种点模式分布下的团伙关系挖掘方法。基于此方法构建团伙关系网络,进而提出一种顾及时空邻近的社团关系度量方法。采用该方法对阿富汗与巴基斯坦地区1970—2018年真实GTD数据进行实验分析,结果表明:该研究区域主要恐怖团伙中共存在11种团伙关系({Taliban,TTP}等),在团伙关系网络中占据核心地位的恐怖组织为Taliban,其次为TTP、BLA。本文提出的方法能有效地挖掘出恐怖团伙之间潜在的关系,团伙关系的定量表达度量了不同组织在关系网络中的地位,验证分析表明,挖掘结果与相应团伙共同参与事件的时空分布、历史根源高度吻合。本文拓展了时空同现模式挖掘在恐怖团伙关系发现中的应用,研究成果对公安部门在识别某些恐怖事件制造者以及合理分配警力资源方面具有重要意义。

关键词: 时空数据, 时空邻近, 同现模式, 恐怖团伙, 点模式, 关系度量, GTD, 关联

Abstract:

The existing methods of mining the relationship between terrorist groups are difficult to be applied in anti-terrorism researches because of the unique characteristics of terrorist group relationship such as complex structure, flexible action, and long survival period. From the perspective of time and space, based on the characteristics of spatiotemporal proximity and spatiotemporal co-occurrence of terrorist groups, the relationship of multiple independently distributed terrorist groups is modeled as the frequency of spatiotemporal co-occurrence in our study. We propose a method of terrorist group relationship mining based on the distribution of point mode. Based on this method, a terrorist group network can be constructed. We further applied this method to analyze the real GTD data of Afghanistan and Pakistan from 1970 to 2018. The results show that there are mainly 11 types ({Taliban,TTP}, etc.) of terrorist group relationship in our study area. The terrorist organization that occupy the core position in the network is Taliban, followed by TTP and BLA. The method proposed in this study can effectively quantify the potential relationship between terrorist groups. The quantified terrorist group relationship also indicates the positions of different organizations in the network. Our verification analysis shows that our results are highly consistent with the spatiotemporal distribution and historical roots of the terrorist events. This study expands the application of spatiotemporal co-occurrence pattern in discovering the relationship of terrorist groups. Our results are of great significance to the public security department in identifying the perpetrators of certain terrorist incidents and reasonably allocating police resources.

Key words: spatiotemporal data, spatiotemporal proximity, co-occurrence pattern, terrorist group, point model, relationship measurement, Global Terrorism Database, association