What is fractional and integral over sampling?

Fractional oversampling refers to the process of generating more samples than the original dataset, but not necessarily an integer multiple, often used to balance class distributions in datasets. It allows for the creation of synthetic examples based on existing data points, commonly through techniques like SMOTE (Synthetic Minority Over-sampling Technique). Integral oversampling, on the other hand, involves creating a whole number of additional samples, typically by duplicating existing data points or generating synthetic samples in whole multiples. Both methods aim to improve model performance, particularly in imbalanced classification problems.