Scholarlink

This research focuses on developing a sophisticated financial agent capable of autonomously learning from past transactions to make optimized trading decisions, ultimately aiming to maximize trading profits. The agent is designed to strategically buy or sell all stocks when market conditions reach perceived peaks or bottoms, or to hold when neither action is ideal. This dynamic decision-making process relies on advanced deep learning algorithms, including Proximal Policy Optimization (PPO), Advantage Actor-Critic (A2C), and Deep Q-Learning (DQN). These algorithms enable the agent to continually refine its performance, improving its trading strategies based on real-time market data and historical insights. Beyond simply automating the trading process, the agent strives to maximize returns even during volatile, high-risk periods in the financial market. We assessed the model’s effectiveness using metrics such as Max Drawdown, Profit and Loss (PnL), and Return on Investment (ROI), achieving promising results of 61.936% for Max Drawdown, 7104.491 in PnL, and 71.044% in ROI, underscoring the agent’s potential to navigate complex financial conditions while maximizing profitability.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Mastering Market Peaks and Valleys: Advanced Trading with Deep Reinforcement Learning

  • 章节

  • 发表日期：

    2026年

查看全文
  • Long Tran Hoang,
  • Hai Nguyen Dinh Vu,
  • Hoa Nguyen Thanh,
  • Anh Nguyen Tran Phuong

摘要

This research focuses on developing a sophisticated financial agent capable of autonomously learning from past transactions to make optimized trading decisions, ultimately aiming to maximize trading profits. The agent is designed to strategically buy or sell all stocks when market conditions reach perceived peaks or bottoms, or to hold when neither action is ideal. This dynamic decision-making process relies on advanced deep learning algorithms, including Proximal Policy Optimization (PPO), Advantage Actor-Critic (A2C), and Deep Q-Learning (DQN). These algorithms enable the agent to continually refine its performance, improving its trading strategies based on real-time market data and historical insights. Beyond simply automating the trading process, the agent strives to maximize returns even during volatile, high-risk periods in the financial market. We assessed the model’s effectiveness using metrics such as Max Drawdown, Profit and Loss (PnL), and Return on Investment (ROI), achieving promising results of 61.936% for Max Drawdown, 7104.491 in PnL, and 71.044% in ROI, underscoring the agent’s potential to navigate complex financial conditions while maximizing profitability.