Beyond global and local multi-target learning

Basgalupp, Márcio; Cerri, Ricardo; Schietgat, Leander; Triguero, Isaac; Vens, Celine

Abstract

In multi-target prediction, an instance has to be classified along multiple target variables at the same time, where each target represents a category or numerical value. There are several strategies to tackle multi-target prediction problems: the local strategy learns a separate model for each target variable independently, while the global strategy learns a single model for all target variables together. Previous studies suggested that the global strategy should be preferred because (1) learning is more efficient, (2) the learned models are more compact, and (3) it overfits much less than the local strategy, as it is harder to overfit on several targets at the same time than on one target. However, it is not clear whether the global strategy exploits correlations between the targets optimally. In this paper, we investigate whether better results can be obtained by learning multiple multi-target models on several partitions of the targets. To answer this question, we first determined alternative partitions using an exhaustive search strategy and a strategy based on a genetic algorithm, and then compared the results of the global and local strategies against these. We used decision trees and random forests as base models. The results show that it is possible to outperform global and local approaches, but finding a good partition without incurring in overfitting remains a challenging task.

Citation

Basgalupp, M., Cerri, R., Schietgat, L., Triguero, I., & Vens, C. (2021). Beyond global and local multi-target learning. Information Sciences, 579, 508-524. https://doi.org/10.1016/j.ins.2021.08.022

• Journal Article Type: Article
• Acceptance Date: Aug 6, 2021
• Online Publication Date: Aug 12, 2021
• Publication Date: 2021-11
• Deposit Date: Sep 2, 2021
• Publicly Available Date: Aug 13, 2022
• Journal: Information Sciences
• Print ISSN: 0020-0255
• Electronic ISSN: 1872-6291
• Publisher: Elsevier
• Peer Reviewed: Peer Reviewed
• Volume: 579
• Pages: 508-524
• DOI: https://doi.org/10.1016/j.ins.2021.08.022
• Keywords: Artificial Intelligence; Information Systems and Management; Computer Science Applications; Theoretical Computer Science; Control and Systems Engineering; Software
• Public URL: https://nottingham-repository.worktribe.com/output/6140075
• Publisher URL: https://www.sciencedirect.com/science/article/abs/pii/S0020025521008227?via%3Dihub
• Additional Information: This article is maintained by: Elsevier; Article Title: Beyond global and local multi-target learning; Journal Title: Information Sciences; CrossRef DOI link to publisher maintained version: https://doi.org/10.1016/j.ins.2021.08.022; Content Type: article; Copyright: Crown Copyright © 2021 Published by Elsevier Inc. All rights reserved.