Portfolio Optimization using Deep Reinforcement Learning
نویسندگان
1 Department of Management, Masjed- Soleiman Branch, Islamic Azad University, Masjed -Soleiman, Iran
2 Department of Industrial Engineering, Masjed -Soleiman Branch, Islamic Azad University, Masjed-Soleiman, Iran
3 Department of management, Masjed-Soleiman Branch, Islamic Azad University, Masjed -Soleiman, Iran
doi
10.22091/jemsc.2025.11158.1192چکیده
This research aims to train an intelligent trader by using artificial intelligence concepts that can help to make optimal decisions for investing in the stock portfolio. For this purpose, a method based on Q deep reinforcement learning is presented for portfolio optimization. In this method, the policy network and the target policy network are used to learn the actions, and the learning network and the target network are used to estimate the optimal Q. The data related to the companies constituting the Dow Jones Industrial Average (DJIA) from March 2008 to October 2021 are used to evaluate the proposed method. Moreover, the performance of the proposed method is compared with conventional investment strategies and two deep reinforcement learning algorithms, PPO and SAC. The results indicate that the proposed method has the best performance on the test data with a total profit of 35.6% compared to other investigated methods. On the other hand, the Sharpe ratio of the proposed method is the highest value, which implies this strategy performs better in balancing profit and risk.
