Releasing anonymized personal data is an important method to promote the use and application of specific data. In statistical disclosure control, anonymized personal data created through various techniques must balance disclosure risk and data utility. Synthetic data is generated data that differs from the original data in value but retains intrinsic information from it. In this paper, we generate synthetic data using a variational autoencoder, a probabilistic graphical model implemented through an artificial neural network. We evaluate data utility, specifically the information loss of the generated synthetic data, and validate the data through regression analysis results.