An Improved Elastic Net for Cancer Classification and Gene Selection

This paper presents an improved elastic net to identify relevant genes for cancer classification. By introducing the data-driven weight coefficients, the improved elastic net can adaptively select genes in groups and reduce the shrinkage bias for the coefficients of significant genes. Moreover, the irrelevant observations on the augmented dataset are removed and the computational complexity is largely reduced. Experiment results on the acute leukaemia data are provided to verify the proposed method.