Abstract

Many political and societal factors can affect the course of geopolitical events, making them extremely challenging to forecast. In this work, we seek to develop a computational approach for forecasting geopolitical events. We show that one can leverage past geopolitical interactions to predict future related geopolitical interactions that have not necessarily occurred before. We propose to leverage ICEWS, a geopolitical interaction event dataset, to develop a predictive modeling framework for geopolitical events using both supervised and unsupervised machine learning approaches.

Date

January 1, 1970

Authors

Mehrnoosh Mirtaheri, Sami Abu-El-Haija, Tozammel Hossain, Fred Morstatter, Aram Galstyan

Journal

ICML Workshop on Learning and Reasoning with Graph-Structured Representations, Long Beach, CA