We investigated data augmentation method for quantitative data following a normal distribution, with a particular focus on fields like medicine and bioinformatics where securing sufficient sample sizes is often challenging. First, we conducted experiments using the widely used Iris dataset. The results showed that data augmentation improved prediction accuracy for both classification and regression problems. Furthermore, we validated the method using high-dimensional, small-sample metabolome data, demonstrating that it achieved a certain level of generalization performance.
SAKAMOTO et al. (Wed,) studied this question.