Abstract
Classification and regression trees, as well as their variants, are off-the-shelf methods in Machine Learning. In this paper, we review recent contributions within the Continuous Optimization and the Mixed-Integer Linear Optimization paradigms to develop novel formulations in this research area. We compare those in terms of the nature of the decision variables and the constraints required, as well as the optimization algorithms proposed. We illustrate how these powerful formulations enhance the flexibility of tree models, being better suited to incorporate desirable properties such as cost-sensitivity, explainability, and fairness, and to deal with complex data, such as functional data.
Original language | English |
---|---|
Journal | TOP |
Volume | 29 |
Issue number | 1 |
Pages (from-to) | 5-33 |
Number of pages | 29 |
ISSN | 1134-5764 |
DOIs | |
Publication status | Published - Apr 2021 |
Bibliographical note
Published online: 17. Marts 2021Keywords
- Classification and Regression Trees
- Tree ensembles
- Mixed-integer linear optimization
- Continuous nonlinear optimization
- Sparsity
- Explainability