• decision tree
• oblique tree
• model transparency
• merging decision trees
• xai
• explainable ai

Decision tree classifiers have been proved to be among the most interpretable models due to their intuitive structure that illustrates decision processes in form of logical rules.
Unfortunately, their interpretability is paid with the simplicity of a learning process that not always returns very accurate models.
More complex tree-based classifiers such as oblique trees and random forests overcome the accuracy of decision trees at the cost of becoming non interpretable.
However, they are still based on logical rules describing the reasons for the decision process.
We propose a single tree approximation method that takes as input any tree-based classifier and returns a single decision tree able to approximate its behavior.
Our proposal is based on two main components: the first one is the merging of tree-based classifiers through an intensional and extensional approach, the second one is the application of a post-hoc explanation strategy.
Moreover, these two components can be combined to reach improved levels of interpretability and accuracy.