Predictive models of world conflict: accounting for regional and conflict-state differences

Abstract

Developing nation-conflict forecasting models is of vital importance to international resource allocation strategies that affect regional and worldwide security and stability. A nation in conflict may evolve very differently to one that is not. Using a unique set of open-source data, a suite of region-specific models are developed that predict nation-state transitions into and out of violent conflict. Conflict transition predictor variables differ not only between the regions, but also within each region dependent on if a nation is identified as in conflict or not. Accurate modeling of complex regional environments is achieved with parsimonious and interpretable models. While we ensure the causal variables of the previous study are considered for inclusion in the models developed, this paper focuses on the ability to accurately predict conflict over individual independent variable analysis.