Empirical Analysis of Assessments Metrics for Multi-class Imbalance Learning on the Back-Propagation Context

摘要

In this paper we study some of the most common assessment metrics employed to measure the classifier performance on the multi-class imbalanced problems. The goal of this paper is empirically analyzing the behavior of these metrics on scenarios where the dataset contains multiple minority and multiple majority classes. The experimental results presented in this paper indicate that the studied metrics might be not appropriate in situations where multiple minority and multiple majority classes exist.