Data complexity assessment in undersampled classification of high-dimensional biomedical data