Predictions for November 2014 -- Country Watchlists

The current version of the CRISP package includes data up until October 2014. Forecasts here are for November 2014.

These hierarchical model forecasts are for the five EOIs (Events of Interest) in the current CRISP package version. Listed in this watch list are countries that did not have an actual incidence of the EOI, but are predicted to be at high risk of having an onset.

实验室文章:迈向冲突预测的新时代

实验室文章:迈向冲突预测的新时代. 2013. International Studies Review.

文章摘要

在冲突研究的领域中,虽然预测分析的重要性不言可喻,但是却一直没有受到足够的重视。我们认为,预测不仅具有实质公共政策参考的能力,另一方面也能用来检证既有理论模型、避免统计上过度配适(overfitting)且降低确认误差(confirmation bias),藉以建构出更可靠的冲突预测。在本篇文章中,我们回顾了学界在冲突预测研究中有哪些进展,发现由于这五十年来学科在资料搜集和运算能力的进步下,研究者得以从事过去所难以企及的预测研究工作,尤其在自动化的编码程序辅助下,快速的搜集数字化的新闻讯息成为可能,冲突研究得以应用以每日、每周、每月为单位的事件解析数据(disaggregated event data)来进行国家层次以下,有关政府与反抗团体的个体活动资料进行及时性的冲突预测工作。

为了呈现冲突研究在过去几年的重大进展,本文重新检视Fearon and Laitin (2003)这份奠定冲突研究基础的文献,从而比较和凸显预测分析在近几年的进展。结果发现,虽然Fearon and Laitin的研究中有很多的解释变量具有统计上的显著性,但是模型对于样本外事件的预测精确度却不高,这因为利用观察型的资料建构出具有统计上显著变量的模型,并无法回答像是何时、何处会发生内战这种决策者所关注的预测问题。

Predictions for May 2014 -- Country Watchlists

The current version of the CRISP package includes data up until April 2014. Forecasts here are for May 2014.

These hierarchical model forecasts are for the five EOIs in the current CRISP package version. Listed countries are either the three with the highest risk for the respective EOI, or all of the countries with a probability of occurrence higher than 0.9. Only countries that did not have an actual incidence of the EOI of interest are listed in the respective watch list.

Pages