ENSEMBLE METHODS FOR CLASS IMBALANCE LEARNING - Imbalanced Learning: Foundations, Algorithms, and Applications - page 59

ENSEMBLE METHODS FOR CLASS
IMBALANCE LEARNING

XU-YING LIU
School of Computer Science and Engineering, Southeast University, Nanjing, China
ZHI-HUA ZHOU
National Key Laboratory for Novel Software Technology, Nanjing University, Nanjing,
China

Abstract: Ensemble learning is an important paradigm in machine learning, which
uses a set of classifiers to make predictions. The generalization ability of an ensem-
ble is generally much stronger than that of individual ensemble members. Ensemble
learning is widely exploited in the literature of class imbalance learning. This
chapter introduces ensemble learning and gives an overview of ensemble methods
for class imbalance learning.

4.1 INTRODUCTION

Ensemble methods use a set of classifiers to make predictions. The generaliza-
tion ability of an ensemble is usually much stronger than that of the individual
ensemble members. Ensemble learning is one of the main learning paradigms in
machine learning and has achieved great success in almost everywhere learning
methods are applicable, such as object detection, face recognition, recommend-
ing systems, medical diagnosis, text categorization, and so on. For example,
an ensemble architecture was proposed by Huang et al. [1] for pose-invariant
face recognition. It trains a set of view-specific neural networks (NNs) and then
uses sophisticated combination rules to form an ensemble. Conventional methods
train a single classifier and require pose information as input, while the ensemble