We use machine learning methods for modeling multi-period corporate default probabilities and obtain higher prediction accuracy compared to linear models with the differences being larger for longer prediction horizons. Overall, tree-boosting has the highest prediction accuracy. In addition, we...