Hybrid Feature Selection Method using Novel Pasi-Luukka and Genetic Algorithm Method for Microarray Cancer Classification

Abstract

Deoxyribonucleic acid (DNA) microarray technology enables the simultaneous measurement of the expression level of numerous genes, thus enabling the identification of patterns in gene expression that may cause a disease or a particular biological process. The DNA microarray technology can identify cancer cells by analysing the gene expression difference between normal cells and cancer cells. However, due to the vast number of features in the DNA microarray, the feature selection method is required to identify the most relevant subset of microarray features for subsequent analysis. In this research paper, a novel hybrid feature selection method called Pasi Luukka-Genetic Algorithm + Support Vector Machine is introduced. This approach combines the strengths of filter and wrapper methods to effectively select features from eight (8) cancer datasets. The Pasi Luukka algorithm filters irrelevant features and reduces dimensionality. A metaheuristic-based feature selection, which is the Genetic Algorithm (GA), selects the optimum features from the filtered features. This paper evaluates the performance of the proposed method, and a comparison work is conducted against other existing hybrid feature selection methods in the literature. The evaluation considers accuracy metrics and the number of selected features using the same microarray datasets.