Rather than compiling real-world records, synthetic datasets are collected using computer programs. Synthetic datasets are made to be flexible and long-lasting enough to help train machine-learning models.