A Survey of Synthetic Data Generation for Machine Learning

Data is the fuel of machine learning algorithms, therefore data generation in machine learning is becoming an important topic. The problem is that finding enough data for machine learning algorithms in some domains or situations is difficult. For example, some data may invade the privacy of people or some other datasets can be related to national security and difficult to be unveiled. This paper reviews the related work in synthetic data generation in terms of available methods for data generation (augmentation) and how the generated data helps in improving the performance of machine learning algorithms. The main focus of this paper is data synthetic methods in the healthcare domain.