Progressive Heterogeneous Ensemble Learning for Cancer Gene Expression Classification

Abstract

Ensemble learning is an extensively researched subject in machine learning due to its robust and reliable performance. Multiple machine learning models are combined in ensemble learning to improve performance and reliability. There are many algorithms and variations in ensemble learning, but most techniques focus on data space like Bagging, AdaBoost, etc., or feature space like Random Subspace, Attribute Bagging, etc. Traditionally an ensemble of the same learning algorithm is used, but to increase the diversity of base classifiers, using an ensemble of different learning algorithms will be more beneficial. Progressive Subspace Ensemble Learning (PSEL) is investigated in this paper, which combines both data sample and feature space at the same time and applies a sequential selection strategy to select the best classifiers. This work extends PSEL as a heterogeneous ensemble of classifiers to improve performance and reliability for cancer gene expression classification. The proposed work is tested for cancer gene expression classification and performs better than state-of-the-art ensemble techniques.