An Approach to Evaluating Learning Algorithms for Decision Trees

TL;DR

This paper introduces a new method to evaluate decision tree learning algorithms by comparing learned models to reference oracles, aiming to improve trustworthiness and reduce costly validation.

Contribution

It presents a novel oracle-centered evaluation approach for decision tree algorithms, involving data generation from reference trees and comparison to learned models.

Findings

Assessed five decision tree algorithms using the proposed method.

The approach effectively estimates the quality of learning algorithms.

Provides a new metric for evaluating decision tree learning ability.

Abstract

Learning algorithms produce software models for realising critical classification tasks. Decision trees models are simpler than other models such as neural network and they are used in various critical domains such as the medical and the aeronautics. Low or unknown learning ability algorithms does not permit us to trust the produced software models, which lead to costly test activities for validating the models and to the waste of learning time in case the models are likely to be faulty due to the learning inability. Methods for evaluating the decision trees learning ability, as well as that for the other models, are needed especially since the testing of the learned models is still a hot topic. We propose a novel oracle-centered approach to evaluate (the learning ability of) learning algorithms for decision trees. It consists of generating data from reference trees playing the role of oracles, producing learned trees with existing learning algorithms, and determining the degree of correctness (DOE) of the learned trees by comparing them with the oracles. The average DOE is used to estimate the quality of the learning algorithm. the We assess five decision tree learning algorithms based on the proposed approach.