Modeling Conflict Duration:

Insights from Ensemble Learning


I investigate why some wars last longer than others. Challenging the dominant rationalist paradigm that champions credible commitment problems as the primary culprit for the longevity of civil wars, I focus on making accurate temporal predictions based on the structural determinants that leverage the maximum variation in data. More specifically, I posit a theory of strategic decision-making under constraints, in a world where absolute and relative capabilities—material, physical, and political—determine the outcome.

My empirical strategy is as follows. First, I replicate 16 studies on conflict duration featuring binary-time-series-cross-section data to identify and compare variable importance measures across studies. Second, using the Cunningham and Lemke (2013) data, which contains features on both interstate and civil wars, I employ ensemble machine learning models using a maximum dissimilarity measure. Finally, I train deep learning models using Keras to capture the complex underlying interactions and make temporal predictions.

Although ensemble learning is a machine learning concept—a committee of weak learners aggregating into a strong learner—I utilize it in two ways in my dissertation. First, I use a diverse set of algorithms, which have their own strengths and weaknesses, to maximize predictive accuracy. But the whole enterprise can be seen as ensemble learning as well—by combining traditional large-n research, machine learning, and a shadow case study, I aim to make a strong empirical case for conceptualizing conflict duration in an unified way.

The dissertation is set to be submitted by July 2018.