Academia.eduAcademia.edu

Figure 2 – uploaded by Neda Yousefi

See full PDFdownloadDownload figure
Evaluation criteria should be used to measure the efficiency of proposed deep reinforcement learning algorithms for stock trading and what the results of such methods will be in practice. The main goal of stock trading is to maximize long-term profit. Usually and the same here, criteria such as annualized return (AR) and sharp ratio (SR) are used to measure and evaluate financial strategies and stock trading performance. The annualized return is the geometric average amount of money earned by an investment each year over a given time period. [16]

Table 2 Evaluation criteria should be used to measure the efficiency of proposed deep reinforcement learning algorithms for stock trading and what the results of such methods will be in practice. The main goal of stock trading is to maximize long-term profit. Usually and the same here, criteria such as annualized return (AR) and sharp ratio (SR) are used to measure and evaluate financial strategies and stock trading performance. The annualized return is the geometric average amount of money earned by an investment each year over a given time period. [16]

Related Figures (8)

Figure 1. Deep Deterministic Policy Gradient (DDPG) architecture
Figure 1. Deep Deterministic Policy Gradient (DDPG) architecture
Figure 3. Advantage Actor Critic (A2C) architecture Figure 2. Deep Deterministic Policy Gradient (DDPG) Algorithm
Figure 3. Advantage Actor Critic (A2C) architecture Figure 2. Deep Deterministic Policy Gradient (DDPG) Algorithm
Figure 5. Stock price history as per close price
Figure 5. Stock price history as per close price
Figure 6. DDPG loss during the learning process in various iterations
Figure 6. DDPG loss during the learning process in various iterations
Figure 8. Convergence of the A2C trading agent in various iterations By comparing two Figure 7, Figure 8, it is obvious that, unlike the DDPG trading agent, almost all the iterations had the same behavior and all converged to a specific value, but about the A2C trading agent, different iterations lead to different results. In addition, the average results from the A2C trading agent are lower than the DDPG trading agent.
Figure 8. Convergence of the A2C trading agent in various iterations By comparing two Figure 7, Figure 8, it is obvious that, unlike the DDPG trading agent, almost all the iterations had the same behavior and all converged to a specific value, but about the A2C trading agent, different iterations lead to different results. In addition, the average results from the A2C trading agent are lower than the DDPG trading agent.
Table 1. Results by applying the test dataset (AR=Annualized Return and SR= Sharpe Ratio)
Table 1. Results by applying the test dataset (AR=Annualized Return and SR= Sharpe Ratio)
Related topics:
Computer ScienceArtificial IntelligenceReinforcement LearningAlgorithmic TradingLearning classifier systemTrading Strategy
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved