[2P5-GS-11] AI and Society

[2P5-GS-11-03] Stock Market Simulation by Micro-Macro GAN

〇Yusuke Naritomi^1,2, Takanori Adachi² (1. Mitsubishi UFJ Trust Investment Technology Institute Co., Ltd., 2. Graduate School of Management, Tokyo Metropolitan University)

[[Online]]

Keywords:Generative Adversarial Network, Stock Market Simulation, Top-down Approach

In recent years, research in the field of finance has begun to use deep learning models for stock market simulations. However, it is fundamentally difficult to generate realistic macro price dynamics from micro order dynamics. In this study, we propose a new market simulation method by using a deep learning model to generate the macro price dynamics from the micro order dynamics. The market simulator built by deep learning model, which is called Micro-Macro GAN, is trained by coupling two mechanisms. The first mechanism generates the micro order dynamics and trains by using the WGAN, which is called Micro GAN. The second mechanism generates the macro price dynamics from the micro order dynamics generated by the Micro GAN and trains by using the Sig-W GAN, which is called the Macro GAN. As a demonstration, training was performed on Toyota Motor Corporation taken from the FLEX Full data. The order dynamics generated from the Micro GAN and the Micro-Macro GAN were compared to the real order dynamics data. The results showed that the Micro-Macro GAN results were more similar to the real dynamics than the Micro GAN results.