Combining One-Class Classifiers via Meta-Learning

TL;DR

This paper introduces TUPSO, a meta-learning based ensemble method for one-class classifiers, demonstrating superior performance and robustness compared to existing ensembles through extensive experiments.

Contribution

It proposes a novel ensemble scheme, TUPSO, and new performance measures for one-class classifiers, advancing ensemble methods in one-class classification tasks.

Findings

TUPSO outperforms other ensemble methods in experiments.

TUPSO's performance is statistically similar to the best possible classifier.

New performance measures effectively weigh classifiers in ensembles.

Abstract

Selecting the best classifier among the available ones is a difficult task, especially when only instances of one class exist. In this work we examine the notion of combining one-class classifiers as an alternative for selecting the best classifier. In particular, we propose two new one-class classification performance measures to weigh classifiers and show that a simple ensemble that implements these measures can outperform the most popular one-class ensembles. Furthermore, we propose a new one-class ensemble scheme, TUPSO, which uses meta-learning to combine one-class classifiers. Our experiments demonstrate the superiority of TUPSO over all other tested ensembles and show that the TUPSO performance is statistically indistinguishable from that of the hypothetical best classifier.