Evolutionary Computation for Optimal Ensemble Classifier in Lymphoma Cancer Classification

Owing to the development of DNA microarray technologies, it is possible to get thousands of expression levels of genes at once. If we make the effective classification system with such acquired data, we can predict the class of new sample, whether it is normal or patient. For the classification system, we can use many feature selection methods and classifiers, but a method cannot be superior to the others absolutely for feature selection or classification. Ensemble classifier has been using to yield improved performance in this situation, but it is almost impossible to get all ensemble results, if there are many feature selection methods and classifiers to be used for ensemble. In this paper, we propose GA based method for searching optimal ensemble of feature-classifier pairs on Lymphoma cancer dataset. We have used two ensemble methods, and GA finds optimal ensemble very efficiently.

• Type: Conference
• Year: 2003
• Where: ISMIS
• Authors: Chanho Park, Sung-Bae Cho
• Keywords: DNA Microarray Technologies, Feature Selection, Feature Selection Methods, ISMIS 2003