Skip to main navigation Skip to search Skip to main content

New optimization models for optimal classification trees

Research output: Contribution to journalArticlepeer-review

Abstract

The most efficient state-of-the-art methods to build classification trees are greedy heuristics (e.g. CART) that may fail to find underlying characteristics in datasets. Recently, exact linear formulations that have shown better accuracy were introduced. However they do not scale up to datasets with more than a few thousands data points. In this paper, we introduce four new formulations for building optimal trees. The first one is a quadratic model based on the well-known formulation of Bertsimas et al. We then propose two different linearizations of this new quadratic model. The last model is an extension to real-valued datasets of a flow-formulation limited to binary datasets (Aghaei et al.). Each model is introduced for both axis-aligned and oblique splits. We further prove that our new formulations have stronger continuous relaxations than existing models. Finally, we present computational results on 22 standard datasets with up to thousands of data points. Our exact models have reduced solution times with learning performances as strong or significantly better than state of the art exact approaches.

Original languageEnglish
Article number106515
JournalComputers and Operations Research
Volume164
DOIs
Publication statusPublished - 1 Apr 2024

Keywords

  • Combinatorial optimization
  • Linearizations
  • Mixed binary programming
  • Optimal classification trees
  • Quadratic programming

Fingerprint

Dive into the research topics of 'New optimization models for optimal classification trees'. Together they form a unique fingerprint.

Cite this