TimeGAN for Extensions of Economic Time Series Data

Due to the recent developments in big data analysis and deep learning, methodologies for quantitatively and scientifically analyzing data are developing, which effectively solve various human social problems such as national, social, and economic.In order to apply such a scientific analysis technique, the quality and quantity of data are essential.Still, when the amount of data is insufficient, it is challenging to apply it to the scientific method.For example, data such as the macroeconomy are often set to have a relatively long period of data disclosure, or because the assets of real estate-related indicators among assets are very non-floating, the presentation period is often long.Therefore, in order to overcome the limitations of this data shortage, this study intends to study how to expand the existing time series through the TimeGAN model.A total of 50 time series related to economic indicators were collected on a monthly basis from 2013 to 2022, according to the data set, and a total of ten final time series were selected through the correlation between time series.As a result of expanding ten time series through the TimeGAN model, the similarity between the statistical distribution of the actual and extended data was proved based on the PCA and t-SNE visualization algorithms.Through this study, the possibility of overcoming various potential limitations arising from narrow data could be confirmed.Therefore, it is expected that this paper will be a positive reference in various future studies to overcome shortcomings such as data underfitting.