Classification and regression tree models are unmatched for their interpretability, a feature that is lacking in "black-box" models, such as tree ensembles and those constructed by deep learning and gradient boosting. Yet tree models have been falling out of favor in recent years. One reason is the prediction accuracy of tree models tends to be lower than that of black-box models, particularly random forests. Consequently, the latter have largely supplanted trees for prediction tasks.
The lower prediction accuracy of tree models versus black-box models had seemed inevitable: you get either interpretability or accuracy, but not both. Current emphasis on "explainable AI" has renewed interest in algorithms that produce single-tree models with predictive accuracy on par with that of black-box models. The lower prediction accuracy tree models stems from their restrictions to (i) splits on a single variable at a time and (ii) a constant prediction in each terminal node. These restrictions were meant to facilitate interpretation. One way to soften these restrictions while retaining explainability and improving accuracy is to construct trees with linear splits and linear models in the nodes. These two ideas have been tried before but never together. They are now implemented in the GUIDE algorithm and software. Empirical evidence based on real data show that these new tree models have predictive accuracy comparable to or better than that of random forests, neural nets, and gradient-boosted trees. The new models can approximately "explain" which variables are utilized and how they are utilized by a black-box model. They can also act as "explainable" replacements for the latter. The GUIDE website is at https://pages.stat.wisc.edu/~loh/guide.html