| 考虑行业指数的深度强化学习投资组合策略 |
| Portfolio strategy via deep reinforcement learning considering industry index |
| 摘要点击 30 全文点击 0 投稿时间：2024-05-21 修订日期：2025-03-06 |
| 查看/发表评论 下载PDF阅读器 |
| 中文关键词 投资组合; 深度强化学习; 行业指数 |
| 英文关键词 portfolio; deep reinforcement learning; industry index |
| 基金项目 国家自然科学基金资助项目(72371080); 广东省基础与应用基础研究基金资助项目(2023A1515012840); 广东省哲学社会科学规划资助项目(GD23XGL022) |
| 作者 | 单位 | 邮编 | | 杨兴雨 | 广东工业大学 | 510520 | | 温嘉林* | 广东工业大学 | 510520 | | 郑萧腾 | 广东工业大学 | | | 陈亮威 | 广东工业大学 | | | 黎嘉豪 | 广东工业大学 | |
|
| 中文摘要 |
| 为了提高投资决策的收益并降低风险，将行业指数引入到深度强化学习框架中设计投资组合策略，使交易智能体在探索的过程中同时关注资产自身的增长潜力和资产所属行业表现的差异。在策略设计过程中，分别构建资产评分网络和行业评分网络，并将资产和行业指数数据输入到两个网络分别得到资产和行业的评估得分，从而得到投资决策。通过在三个真实股票市场数据集上进行数值实验，验证了所设计的投资组合策略在多个评估指标上优于对比策略。研究结果表明，考虑资产增长潜力以及行业表现有助于投资策略从多个行业中准确识别投资机会，适时地选择前景更好的资产，从而增强了策略平衡收益和风险的能力。 |
| 英文摘要 |
| To improve the returns of investment decisions and reduce the risks, this paper incorporates industry index into the deep reinforcement learning framework to design portfolio strategy, which makes the trading agent simultaneously consider the growth potential of assets and the differences in the performance of industries to which the assets belong during the exploring process. In the process of strategy designing, we construct an asset scoring network and an industry scoring network separately. The asset and industry index data are input to the two networks to obtain the assessment scores of assets and industries respectively, and thus the investment decision is obtained. Through numerical experiments on three real stock market datasets, it is validated that the strategy outperforms comparison strategies in multiple evaluation metrics. The results show that considering the growth potential of assets and industry performances can help the investment strategy accurately identify investment opportunities from multiple industries and timely select assets with better prospects, thereby enhancing the strategy's ability to balance returns and risks. |
| 关闭 |
