一种基于部分可观察马尔可夫决策过程的股票交易策略

A Stock Trading Strategy Based on Partially Observable Markov Decision Process

近年来涌现了许多把深度强化学习应用到股票交易策略的研究。深度强化学习通常依赖于马尔可夫决策过程建模,但是股票市场中交易策略的制定需要考虑历史交易数据中包含的信息。因此,本文通过部分可观察马尔可夫决策过程对股票市场建模,并采用长短期记忆网络和优势演员评论家算法来构建股票交易策略。通过在道琼斯工业平均指数成份股数据集上进行实验,实验结果表明本文所设计的股票交易策略构建方法可以挖掘隐藏在历史数据中的有效信息,获得稳定且有效的交易策略。

In recent years, many researches have emerged that apply deep reinforcement learning to stock trading strategies. Deep reinforcement learning is usually based on Markov decision process, but it should consider the information contained in historical data to make a trading strategy in the stock market. Therefore, this paper models the stock market by the partially observable Markov decision process, and uses the long and short-term memory network and the advantage actor critic algorithm to construct the stock trading strategy. Through experiments on the data set of Dow Jones Industrial Average constituent stocks, the results show that the method in this paper can get the effective information in hidden historical data and obtain a stable and effective trading strategy.