非洲恐怖袭击时空规律的大数据分析——基于GIS技术和分离总体持续期模型  

Spatial-Temporal Patterns of Terrorist Attacks in Africa:A Big-Data Analysis with GIS Techniques and Split-Population Duration Modeling

作　　者：陈冲 庞珣[1] CHEN Chong;PANG Xun

机构地区：[1]清华大学国际关系学系,北京100084

出　　处：《外交评论（外交学院学报）》2020年第2期121-154,I0005,共35页Foreign Affairs Review

基　　金：国家社科基金重大项目“地缘政治风险预测的理论与方法研究”(项目编号:17ZDA110)的阶段性成果。

摘　　要：大数据及其分析技术的发展极大地推动了国际关系研究的发展,使得传统国际关系研究无法完成的任务变为可能。恐怖主义是当今全球治理中重大而棘手的问题,也是国际关系领域长期关注的议题。本文运用海量事件数据和地理信息系统(GIS)技术,将恐怖袭击的地理单元细粒化到以0.5×0.5经纬度作为分辨率的时空网格,寻找微观分析单元上恐怖袭击的时空规律,并运用分离总体持续期模型对恐怖袭击风险的空间分布和发生时机进行解释和预测。通过对非洲10674个时空网格在1994-2013年间恐怖袭击的时空特征进行分析,本文对非洲恐怖袭击发生时机进行精准捕捉,对恐袭地点进行细粒化定位,并对恐袭地理分布进行"疫区"与"免疫区"区分,不但深化了对恐怖主义的认识,而且提高了恐袭预测的精度和实际应用价值。本文的研究尝试表明,在大数据时代进行国际关系研究,不仅要充分利用海量数据提供的前所未有的丰富信息,更要选择和运用恰当的统计技术对复杂时空相关性进行细粒度分析,从而形成有效的创新路径。Big Data and related analytical techniques have dramatically expanded the scope of international relations studies and enabled IR scholars to tackle problems that were viewed to be intractable in the past.Terrorism is among the most important issues of global insecurity,and efforts have been made for long to generate and cumulate information and knowledge thereof.Based on massive event data,this article applies Geographical Information System techniques to conduct a fine-grained analysis of researcher-defined geographical units,using the PRIO-GRID vector grid network with a resolution of 0.5 x0.5 decimal degrees.We attempt to reveal the spatial-temporal patterns of terrorist attacks in Africa,consisting of the 10647 grids between 1994 and 2013.We utilize machine-learning methods to build a Split-Population Duration model to locate,explain and predict the risk and timing of terrorist attacks as well as the immunity of each grid to such attacks.This article contributes to the understanding of terrorism and provides a pilot analysis for future efforts to build a counter-terrorism early warning system.We also try to demonstrate that innovations in Big Data IR studies shall rely not only on rich information in big data,but also on appropriate analytical tools that are crucial for a fine-grained analysis of complex relationships in a high-dimensional space and a dynamic process to"turn mission impossible into mission possible".

关 键 词：大数据 恐怖主义 非洲 地理信息系统 时空网格 国际关系研究 事件数据 

分 类 号：D815.5[政治法律—国际关系]