Cancer classification using ensemble of error correcting output codes

Abstract

We address the microarray dataset based cancer classification problem using a newly proposed ensemble of Error Correcting Output Codes (E-ECOC) method. To the best of our knowledge, it is the first time that ECOC based ensemble has been applied to the microarray dataset classification. Different feature subsets are generated from datasets as inputs for some problem-dependent ECOC coding methods, so as to produce diverse ECOC coding matrixes. Then, the mutual difference degree among the coding matrixes is calculated as an indicator to select coding matrixes with maximum difference. Local difference maximum selection(L-DMS) and global difference maximum selection(G-DMS) are the strategies for picking coding matrixes based on same or different ECOC algorithms. In the experiments, it can be found that E-ECOC algorithm outperforms the individual ECOC and effectively solves the microarray classification problem. © 2014 Springer International Publishing Switzerland.