Forecasting conflict on a fine-grained grid level has real-life policy implications that can empirically inform meaningful healthcare and peace-preservation decisions. In many settings, interpretable models have the appeal that policymakers know how to draw conclusions from the model and do not have to base their decisions on black-box models. On the other hand, machine learning methods, such as random forest or neuronal networks, are known to have better prediction performance, begging the question of whether such methods should be preferred over classic models in critical situations that can save lives. Building on the work of Fritz et al. (2022), this project should investigate whether using state-of-the-art machine learning methods are better at predicting state-based conflict in Africa than ordinary logistic regression. A working knowledge of the programming languages R or Python and statistical regression models is strongly advised to work on this project. All interested students must first email fritzc@tcd.ie to set up a short informal meeting to discuss the project.
Relevant literature:
Fritz, Mehrl, Thurner, & Kauermann (2022). The role of governmental weapons procurements in forecasting monthly fatalities in intrastate conflicts: A semiparametric hierarchical hurdle model. International Interactions 48 (4), 778-799