Research And Application Of Stock Quantitative Trading Based On Deep Reinforcement Learning

With the development of computer-related technology,various industries have changed on the basis of tradition,include the financial market.Financial transaction itself is a complicated process,in which massive amounts of data are generated.When faced with these data,the traditional method of calculating and judging based on traders has great limitations.In addition,compared with computers,human have a lot of emotional factors,which are often caused by their own emotional problems,which lead to errors in the key analysis.Today,with the rapid development of AI(Artificial Intelligence)technology,numerous powerful data processing algorithms have been emerged.Most Internet companies have also invested in research on smart finance.Based on the current development of the financial industry,this article introduces deep reinforcement learning algorithms into the field of stock trading to build an intelligent trading model,in order to discover the rules of the market during the learning of massive data,so as to conduct effective transactions and effectively avoid risks.At the same time,it can increase the income of investors.The research work of this article is mainly divided into three parts.First,based on the studying of relevant domestic and foreign documents,the shortcomings of the existing research is summarized,and based on the actual situation of the stock market,a new deep reinforcement learning modeling method on the basis of the previous is provided and improved.Abstraction of state,action,and reward function are studied.Secondly,on the basis of the traditional DQN algorithm,corresponding to actual needs,the RB?DRL deep reinforcement learning algorithm model is proposed to improve the network structure and connection mode,and at the same time the loss function of the network is redefined.The improved model shows good results in the subsequent multiple sets of comparative experiments.Finally,in order to apply the strategy model to practice,a prototype stock quantitative trading system based on deep reinforcement learning has been designed and developed,combined the three parts of model,strategy,and data,the relevant information is displayed to users in the form of web pages for the sake of facilitating their use.In summary,this article is committed to using deep reinforcement learning algorithms to find the trading rules of the financial market in the massive financial data,and to provide investors with a more robust trading strategy.It has certain uniqueness and superiority in comparison with traditional strategies,and has important value of research and practical application.