Comparing Multiobjective Evolutionary Algorithms for Cancer Data Microarray Feature Selection

JULIETA SOL DUSSAUT; PABLO JAVIER VIDAL; IGNACIO PONZONI; ANA CAROLINA OLIVERA

Río de Janeiro

Congreso; IEEE Congress on Evolutionary Computation (IEEE CEC 2018); 2018

Pontificia Universidad Católica de Río de Janeiro

Microarray analysis is gradually becoming an important tool for diagnosis and classification of human cancers. Microarray data consists of thousands of features most of which are irrelevant for classifying microarray gene expression patterns. The election of a minimal subset of features for classification is a challenging task. In this work, a deep analysis and comparison of multiobjective evolutionary algorithms (MOEAs) for Feature Selection of cancer microarray dataset is presented. The experiments are carried out on benchmark gene expression datasets, i.e., Colon, Lymphoma, and Leukaemia available in the literature. A preprocessing of microarray data is carried out in order to remove strongly correlated features. A detailed comparative study is made to analyze the results of the different MOEAs.