Document Type : Research Paper
Authors
1
M.A. Student, Department of Financial Management, Faculty of Management and Economics, SR.C., Islamic Azad University, Tehran, Iran
2
Associate Professor of Agricultural Economics, Department of Economics, Faculty of Economics and Accounting, CT.C., Islamic Azad University, Tehran, Iran
3
Associate professor, Department of Financial Management, Accounting and Financial Engineering, Faculty of Management and Economics, SR.C., Islamic Azad University, Tehran, Iran
10.22075/ijnaa.2025.37619.5455
Abstract
In recent years, the rapid growth of technology and data-driven financial markets has led to an increased interest in using artificial intelligence, and especially reinforcement learning, in designing automated trading systems. In this study, an automated trading system based on reinforcement learning for the stock market is designed and implemented. The main objective of this study is to investigate the efficiency of reinforcement learning models including Proximal Policy Optimization (PPO), Advantage Actor-Critic (A2C), and Deep Q-Network (DQN) compared to traditional investment methods. To train the models, real stock market data and technical indicators are used, and two different reward functions are proposed to evaluate the system performance. The performance of the models is analyzed with criteria such as total return and Sharpe ratio. Experimental results show that reinforcement learning models can provide higher returns than traditional methods in certain conditions, and choosing the appropriate reward function plays a significant role in improving trading strategies. This research, while emphasizing the importance of carefully designing the reward function and selecting appropriate features, offers new perspectives for the development of intelligent systems in financial markets.
Keywords
Subjects
International Journal of Nonlinear Analysis and Applications