First order alternating decision trees
Anneleen Van Assche, Darek Krzywania, Joris Vaneyghen, Jan Struyf, Hendrik Blockeel · Lirias · 2003
An alternating decision tree is a model that generalizes decision trees and boosted decision trees. The main advantages of alternating decision trees over boosting regular decision trees are that they are easier to interpret for human experts and that they can be built efficiently. In this work, we introduce first order alternating decision trees (FOADTs), an upgrade of alternating decision trees to first order logic. We propose three important efficiency improvements, which make inducing FOADTs computationally feasible. Furthermore, we describe some issues with FOADTs that are specific to the first order setting. We conclude with some preliminary experiments.