Abstract: Conflict research has seen a growing emphasis on prediction and forecasting models. We describe automated machine learning (autoML) to identify benchmark models, and human-guided machine learning (HGML) to incorporate domain knowledge and research requirements into model selection and assessment. We analyze three published models of conflict (Goldstone et al. 2010, Gleditsch and Ward 2013, Gelpi and Avdan 2018) to identify routine tasks in predictive modeling. We input each model to our autoML system and find comparable or better models can be selected, using the same data. Our research has two takeaways for conflict modeling: one, predictive models of conflict would benefit from even minimal applications of autoML; two, HGML offers the attractive option of constraining autoML systems to address the kinds of questions conflict researchers assess with predictive models.

Bio: Vito D'Orazio is an Assistant Professor of Political Science at the University of Texas at Dallas. He studies conflict and political violence, with an emphasis on methods to advance conflict modeling. His work has been funded by DARPA and the NSF, and has appeared in outlets such as Political Analysis and IEEE Big Data.