基于深度确定性策略梯度算法的股票投资组合策略研究  

Deep deterministic policy gradient for stock portfolio management

作　　者：董小刚 韩元元 秦喜文 DONG Xiaogang;HAN Yuanyuan;QIN Xiwen(School of Mathematics and Statistics,Changchun University of Technology,Changchun 130012,China;Jilin Provincial Key Laboratory of Data Science and Intelligent Decision,Changchun 130012,China)

机构地区：[1]长春工业大学数学与统计学院,吉林长春130012 [2]吉林省数据科学与智能决策重点实验室,吉林长春130012

出　　处：《东北师大学报(自然科学版)》2025年第1期29-34,共6页Journal of Northeast Normal University(Natural Science Edition)

基　　金：吉林省自然科学基金资助项目(20210101149JC);国家社会科学基金资助项目(2024BTJ074,2023BTJ047)。

摘　　要：为构建更加全面有效的投资组合,采用了深度确定性策略梯度算法,并在奖励函数中引入了风险衡量指标索提诺比率来实现风险与收益之间的权衡.除基本的股票数据外还将股票市场中的技术指标作为状态的输入,以捕捉股票市场的主要趋势.经数据检验,与其他强化学习算法对比,改进奖励函数的DDPG算法能够在控制风险的同时得到较高收益,有效地实现了风险的分散和投资组合的稳健性.In order to construct a more comprehensive and effective portfolio,the Deep Deterministic Policy Gradient(DDPG)algorithm is employed,and the risk-adjusted measure,the Sortino Ratio(SR)is introduced into the reward function to achieve a balance between risk and return.Additionally,besides utilizing fundamental stock data,technical indicators from the stock market are incorporated as inputs to capture the major market trends.To validate the effectiveness of this approach,experiments are conducted in the Chinese A-share market.The results demonstrate that,compared to other reinforcement learning algorithms,higher returns are achieved with the enhanced reward function of the DDPG algorithm,while risk diversification and portfolio robustness are effectively achieved.

关 键 词：股票投资组合 深度强化学习 索提诺比率 深度确定性策略梯度 

分 类 号：O212[理学—概率论与数理统计]