Work in Progress
For a more recent list of work in progress, please see my new website at the University of Konstanz.
Conflict Prediction via Machine Learning: Addressing the Rare Events Problem with Bagging
Poster presented at PolMeth XXV, 25th Annual Summer Conference of the Society for Political Methodology, 9–12 July 2008, University of Michigan.
Abstract: Machine learning deals with the development of algorithms for classification and prediction. However, these algorithms have only in rare cases been used in political science. This poster demonstrates the application of state-of-the-art machine learning techniques to the prediction of conflict. In order to address the rare events problem, I use an ensemble of classifiers built on subsets of the training data. These subsets include all positive cases, and a random selection of negative ones. Although I focus primarily on decision tree learning, the proposed method can be used in conjunction with different classification algorithms in order to improve the prediction of conflict onset.