Machine learning can facilitate data fabrication by generating realistic and intricate synthetic data through the analysis of patterns in existing datasets. This capability enables the simulation of scenarios, testing of edge cases, and augmentation of scarce real-world data for model training, especially in contexts where obtaining authentic data is challenging or costly. Machine learning models may generate synthetic datasets that accurately replicate real-world data distributions, allowing developers to evaluate their algorithms under various settings without jeopardizing sensitive information. Machine learning can augment the training data quantity and enhance model robustness by producing supplementary variations of existing data, particularly in scenarios involving restricted datasets. Machine learning may generate data that simulates unusual events or extreme conditions not easily found in real-world datasets, so aiding in scenario design and risk assessment. In the context of partial datasets, machine learning can be employed to generate missing data points by leveraging patterns identified in the available data. This chapter presents an in depth study of machine learning techniques in data fabrication. This chapter has a special emphasis on data augmentation for medical image classification.