Gene Expression Programming Ensemble for Classifying Big Datasets

Abstract

The paper proposes a new GEP-based batch ensemble classifier constructed using the stacked generalization concept. In our approach combination of base classifiers involves evolving the meta-gene using genes induced by GEP from randomly generated combinations of instances with randomly selected subsets of attributes. The main property of the discussed classifier is its scalability allowing adaptation to the size of the dataset under consideration. To validate the proposed classifier, we have carried-out computational experiment involving a number of publicly available benchmark datasets. Experiment results show that the approach assures good performance, scalability and robustness.