Ensemble-based Feature Selection using Symmetric Uncertainty and SVM classification

Feature selection plays a significant role in gene expression data due to its dimensionality. Microarray gene expression data is particularly important and significant in clinical diagnosis. Our aim is to improve the feature selection technique using the ensembling method. The ranked subset from three filter methods namely spearman, Pearson, MI is used in the EFS-SU algorithm. Symmetric Uncertainty is used to find C-Correlation and F-Correlation. This algorithm will help the SVM classifier in reducing computational cost and finding relevant and efficient genes, without redundancy. Due to the elimination of redundant and non-relevant genes,we were able to increase the classifier accuracy.