[1]宋 飞,田辰磊,冯传威.基于深度强化学习的股票量化投资研究[J].南京师大学报(自然科学版),2025,48(01):85-92.[doi:10.3969/j.issn.1001-4616.2025.01.011]
　Song Fei,Tian Chenlei,Feng Chuanwei.Research of Stock Quantitative Trading Algorithm Based on Deep Reinforcement Learning[J].Journal of Nanjing Normal University(Natural Science Edition),2025,48(01):85-92.[doi:10.3969/j.issn.1001-4616.2025.01.011]
点击复制

基于深度强化学习的股票量化投资研究()
分享到：

《南京师大学报（自然科学版）》[ISSN:1001-4616/CN:32-1239/N]

卷:
48
期数:
2025年01期
页码:
85-92
栏目:
计算机科学与技术
出版日期:
2025-02-15

文章信息/Info

Title:
Research of Stock Quantitative Trading Algorithm Based on Deep Reinforcement Learning
文章编号:
1001-4616(2025)01-0085-08
作者:
宋 飞1; 2; 田辰磊2; 冯传威3
(1.南京大学信息管理学院,江苏 南京 210023)
(2.南京林业大学理学院,江苏 南京 210037)
(3.南京大学数学学院,江苏 南京 210093)
Author(s):
Song Fei1; 2; Tian Chenlei2; Feng Chuanwei3
(1.School of Information Management,Nanjing University,Nanjing 210023,China)
(2.College of Science,Nanjing Forestry University,Nanjing 210037,China)
(3.School of Mathematics,Nanjing University,Nanjing 210093,China)
关键词:
量化投资; DQN; 强化学习; 算法交易
Keywords:
quantitative investment; DQN; reinforcement learning; algorithmic trading
分类号:
TP18
DOI:
10.3969/j.issn.1001-4616.2025.01.011
文献标志码:
A
摘要:
针对股票量化投资,将深度强化学习中的Deep Q-learning(DQN)模型应用于算法交易,构建端到端的算法交易系统. 首先,利用股票技术分析指标设计股票交易环境,从时间尺度扩充特征集; 其次,定义智能体交易的奖励函数和动作空间; 然后,设计Q网络结构,将支持向量机和极致梯度提升法学习股票历史数据的涨跌信号加入强化学习中; 最后,将算法交易系统应用于中国股票市场,并选择招商银行和泰和科技两支股票以及其余4支股票进行验证,从收益率、夏普比率和最大回撤率三方面评价投资绩效,结果表明该算法系统在收益率上有显著提升的同时,最大回撤率有所降低,模型的抗风险能力较高.
Abstract:
This paper focuses on quantitative stock investment and applies the Deep Q-learning(DQN)model from deep reinforcement learning to algorithmic trading,constructing an end-to-end algorithmic trading system. Firstly,the system utilizes stock technology analysis index to design a stock trading environment,expanding the feature set from a time scale; secondly,it defines the reward function and action space for intelligent agent transactions; then,it designs Q-network structure and incorporate Support Vector Machine and eXtreme Gradient Boosting method to learn the rise and fall signals of stock historical data into reinforcement learning; finally,the algorithmic trading system is applied to the Chinese stock market and China Merchants Bank and Taihe Technology,as well as the remaining four stocks are selected for validation. The investment performance is evaluated from three aspects:return rate,sharpe ratio,and maximum drawdown rate. The results are shown that the algorithmic system significantly improved the return rate while reducing the maximum drawdown rate,indicating that the model has a high risk resistance ability.

参考文献/References:

[1]姜宇薇. 金融市场量化交易的国际经验[J]. 中国货币市场,2022,246(4):22-25.
[2]张晓燕,张远远. 量化投资在中国的发展及影响分析[J]. 清华金融评论,2022(1):44-45.
[3]梁天新,杨小平,王良,等. 基于强化学习的金融交易系统研究与发展[J]. 软件学报,2019,30(3):20.
[4]JIANG Z,XU D,LIANG J. A deep reinforcement learning framework for the financial portfolio management problem[J/OL]. arXiv Preprint arXiv:1706.10059,2017.
[5]XIONG Z,LIU X Y,SHAN Z,et al. Practical deep reinforcement learning approach for stock trading[J/OL]. arXiv Preprint arXiv:1811.07522,2018.
[6]韩道岐,张钧垚,周玉航,等. 基于深度强化学习的股市操盘手模型研究[J]. 计算机工程与应用,2020,56(21):145-153.
[7]AZHIKODAN A R,BHAT A,JADHAV M V. Stock trading bot using deep reinforcement learning[M]. Singapore:Innovations in Computer Science and Engineering,2019:41-49.
[8]HUANG Z,LIN N,MEI W L,et al. Algorithmic trading using combinational rule vector and deep reinforcement learning[J]. Applied soft computing,2023,147:110802.
[9]MAHDI M,MAHOOTCHI M. A deep Q-learning based algorithmic trading system for commodity futures markets[J]. Expert systems with applications,2024,237:121711.
[10]TRELEAVEN P,GALAS M,LALCHAND V. Algorithmic trading review[J]. Communications of the ACM,2013,56(11):76-85.
[11]MNIH V,KAVUKCUOGLU K,SILVER D,et al. Playing atari with deep reinforcement learning[EB/OL](2013-12-19)[2023-01-08].
[12]BANOTHS P R,DONTA P K,AMGOTH T. Dynamic mobile charger scheduling with partial charging strategy for WSNs using deep-Q-networks[J]. Neural computing and applications,2021,33(22):15267-15279.
[13]石泽宇. 我国股票市价变动影响因素的实证分析[J]. 北方经贸,2020(9):125-127.
[14]WANG Y,WANG D,ZHANG S,et al. Deep Q-trading[R]. CSLT Technical Report-20160036,2017.
[15]李嘉浩. 基于支持向量机的股票预测与分析[J]. 经济研究导刊,2021(32):107-110.
[16]葛橹漠,周显. 基于XGBoost的多因子选股模型[J]. 信息技术与标准化,2020(5):36-41.

备注/Memo

备注/Memo:
收稿日期:2024-10-10.
基金项目:国家自然科学基金项目(12201303).
通讯作者:宋飞,博士,副教授,研究方向:深度学习. E-mail:songfei@njfu.edu.cn
更新日期/Last Update: 2025-02-15