基于LSTM预测信息的在线融资融券组合交易策略  

Online margin trading strategy based on LSTM prediction information

作　　者：刘伟龙 张永[1] 杨兴雨[1] LIU Weilong;ZHANG Yong;YANG Xingyu(School of Management,Guangdong University of Technology,Guangzhou 510520,China)

机构地区：[1]广东工业大学管理学院,广州510520

出　　处：《系统工程理论与实践》2024年第8期2493-2508,共16页Systems Engineering-Theory & Practice

基　　金：国家自然科学基金面上项目(72371080);教育部人文社会科学研究基金(21YJA630117);广东省基础与应用基础研究基金(2024A1515012670,2023A1515012840);广东省哲学社会科学规划项目(GD23XGL022)。

摘　　要：随着国内融资融券市场日渐成熟和金融科技不断进步,构建智能化的融资融券交易策略成为量化金融领域的重要话题和关键挑战.该文利用长短期记忆网络(LSTM)预测信息构建专家策略,提出了集成专家意见的在线融资融券组合交易策略.首先,使用多个技术指标作为输入变量,通过LSTM神经网络模型预测股价的涨跌趋势.其次,考虑投资于单支股票的专家,根据LSTM模型的预测结果构建各专家的买卖策略.然后,提出了一种基于专家表现的权重优化模型,通过求解模型确定每个专家的权重.最后,为说明所构建策略的有效性,利用股票市场历史交易数据进行实证分析.结果表明所构建的策略能够获得高于基准策略的绩效表现,并在考虑交易费用的情况下仍能保持优越性.With the growing maturity of the domestic margin market and the continuous advancement of financial technology,the development of intelligent margin trading strategies has emerged as a critical topic and challenge in the field of quantitative finance.This article employs long short-term memory(LSTM)networks to construct expert strategies based on predictive information and proposes an online margin portfolio trading strategy that integrates expert opinions.Firstly,multiple technical indicators are utilized as input variables to anticipate the trend of stock prices using the LSTM neural network model.Next,an expert specializing in investing in a single stock is considered,and buying or selling strategies for each expert are formulated based on the LSTM model’s prediction results.Then,a weight optimization model based on expert performance is proposed to determine the weight of each expert by solving the model.Finally,to demonstrate the efficacy of the proposed strategy,historical trading data from stock markets is utilized for empirical analysis.The results demonstrate that the developed strategy is capable of achieving better performance than some benchmark strategies,even when taking transaction costs into consideration.

关 键 词：机器学习 融资融券 在线投资组合 LSTM神经网络 

分 类 号：F832.5[经济管理—金融学]