Random forests can be used to effectively model and estimate discrete choices, providing an interpretable and flexible approach that outperforms existing methods.
