Paper Title
Multi-Domain Generative Adversarial Network for Time-Series Data

Abstract
Time-series data generally refers to information with a time dimension, such as weather data, expenditure records, and even stock trends. These different types of data are pervasive in our lives and profoundly influence our behavior. Predicting time-series data is also an essential task in the field of deep learning. However, time-series data must be recorded over time, which means it is challenging to collect sufficient data within a short period. The lack of information is also a significant problem in the field of deep learning. Therefore, we designed a framework based on a Time-Series Generative Adversarial Network. This method constrains a variety of fields to ensure the stability of the model, reduce the impact of model collapse, and improve the quality of the generated data. The experimental results show that by training simple classification and prediction models, our method outperforms models without data augmentation and models with general data augmentation. Keywords - Time Series, Generative Adversarial Network, Stability.