Gene selection and classification of pancreatic microarray datasets

Abstract

Microarray data is measured with numerous features, selection of the most important genes from such data sets is an essential step towards maximizing classification accuracy. Machine learning algorithms have proved to be the most effective choice in gene selection and cancer classification research. In the past, human cancer classification was basically morphological, introduction of microarray technology has supported simultaneous and efficient analysis of thousands of cancer genes. We have applied combinations of a number of feature selection and classification machine learning algorithms in order to maximize classification accuracy and we have carried out differential expression analyses on the selected genes. Our proposed algorithms combinations yield better performance results than existing literature on these pancreatic datasets.