Back Coup d'Future: An Ensemble Machine Learning Framework for Coups Prediction
The quantitative study of coups has benefited from immense explanatory research, but less so with regard to prediction. I present a systematic framework for modeling coups utilizing a machine learning ensemble approach. Forecasts are assessed at the country level with yearly temporal resolutions. I outline a methodology that covers feature identification, processing and selection, model development and deployment, and finally metrics for assessing the predictive capacity. Forecasts indicate that in 2020 the five most likely countries for a military coup include Thailand, Tunisia, Somalia, Burundi and Ethiopia.