Improving the Validity of Decision Trees as Explanations
This addresses the issue of explainability validity in decision trees for classification and forecasting with tabular data, which is important for ensuring fairness in subgroups, though it is incremental as it builds on existing tree-based methods.
The paper tackles the problem of decision trees providing misleading explanations due to leaves with unbalanced accuracy, which can be interpreted as unfairness among subgroups. They propose training a shallow tree to minimize the maximum misclassification error across leaves, achieving statistical performance comparable to state-of-the-art methods like XGBoost by extending leaves with further models.
In classification and forecasting with tabular data, one often utilizes tree-based models. Those can be competitive with deep neural networks on tabular data and, under some conditions, explainable. The explainability depends on the depth of the tree and the accuracy in each leaf of the tree. We point out that decision trees containing leaves with unbalanced accuracy can provide misleading explanations. Low-accuracy leaves give less valid explanations, which could be interpreted as unfairness among subgroups utilizing these explanations. Here, we train a shallow tree with the objective of minimizing the maximum misclassification error across all leaf nodes. The shallow tree provides a global explanation, while the overall statistical performance of the shallow tree can become comparable to state-of-the-art methods (e.g., well-tuned XGBoost) by extending the leaves with further models.