TY - JOUR

T1 - An Experimental Evaluation of Mixup Regression Forests

AU - Rodriguez, Juan

AU - Juez-Gil, Mario

AU - Arnaiz-Gonzalez, Alvar

AU - Kuncheva, Ludmila

PY - 2020/8/1

Y1 - 2020/8/1

N2 - Over the past few decades, the remarkable prediction capabilities of ensemble methods have been used within a wide range of applications. Maximization of base-model ensemble accuracy and diversity are the keys to the heightened performance of these methods. One way to achieve diversity for training the base models is to generate articial/synthetic instances for their incorporation with the original instances. Recently, the mixup method was proposed for improving the classication power of deep neural networks (Zhang et al., 2017). Mixup method generates articial instances by combining pairs of instances and their labels, these new instances are used for training the neural networks promoting its regularization. In this paper, new regression tree ensembles trained with mixup, which we will refer to as Mixup Regression Forest, are presented and tested. The experimental study with 61 datasets showed that the mixup approach improved the results of both Random Forest and Rotation Forest.

AB - Over the past few decades, the remarkable prediction capabilities of ensemble methods have been used within a wide range of applications. Maximization of base-model ensemble accuracy and diversity are the keys to the heightened performance of these methods. One way to achieve diversity for training the base models is to generate articial/synthetic instances for their incorporation with the original instances. Recently, the mixup method was proposed for improving the classication power of deep neural networks (Zhang et al., 2017). Mixup method generates articial instances by combining pairs of instances and their labels, these new instances are used for training the neural networks promoting its regularization. In this paper, new regression tree ensembles trained with mixup, which we will refer to as Mixup Regression Forest, are presented and tested. The experimental study with 61 datasets showed that the mixup approach improved the results of both Random Forest and Rotation Forest.

KW - Mixup

KW - Random forest

KW - Regression

KW - Rotation forest

U2 - https://doi.org/10.1016/j.eswa.2020.113376

DO - https://doi.org/10.1016/j.eswa.2020.113376

M3 - Article

VL - 151

JO - Expert Systems with Applications

JF - Expert Systems with Applications

SN - 0957-4174

M1 - 113376

ER -