CFSES Optimization Feature Selection with Neural Network Classification for Microarray Data Analysis

The DNA microarray data are enabling to measure the countenance levels of genes simultaneously, so long as an excessive for tuitous for cancer prediction. The number of features or genes is often more than thousands but in another side, the number of instances or subjects is less than hundreds. For that, it is required and vital to perform gene selection for classification. Feature selection moderates this problem by eradicating irrelevant genes or features from data. So it optimizes the computational cost with concentrated data. At here, we propose a new methodology for feature selection to determine marker genes that are relevant to a type of cancer. A different bio-stimulated procedure entitled as CFSES (Correlation-based Feature Selection Elephant Search) is projected in this work. We use the Linear Support Vector Machine (Linear-SVM), SMO-SVM and MLP to make classification with use of the selected marker genes. The performance of marker gene selection and classification are illustrated in two real data analysis.