Beyond Point Prediction: Artificial Representative Trees with Uncertainty

This paper addresses the trade-off between prediction accuracy and interpretability in machine learning. Random forests achieve strong predictive performance but lack transparency, while individual decision trees offer interpretability but suffer from instability. The authors develop Artificial Representative Trees as interpretable surrogate models for random forests and combine them with leaf-wise Mondrian Conformal Predictive Systems to enable a single tree to provide continuous predictions, prediction intervals, and probabilistic outputs with rigorous coverage guarantees.