Machine learning as a tool in political science

Abstract

The use of machine learning is rapidly gaining ground in empirical political science and public policy making. Machine learning can be employed to predict individual-level outcomes and thus holds potential for increasing precision in targeted early interventions across various policy domains. This article introduces machine learning as a part of the political science and public policy toolbox. It explains key concepts and outlines how machine learning can be carried out in practice. A decision tree model used to predict dropout among students at Copenhagen University College serves as an illustrative case throughout the article. Lastly, the article discusses some of the methodological promises and pitfalls of using machine learning in a social science setting.