Study On The Application Of Multi-agent Rein-forcement Learning Model In A Stock Timing And Stock Selection

In recent years,with the application of artificial intelligence technology in the financial field,its application continues to extend.Among them,deep reinforcement learning algorithm(DRL)is more prominent in artificial intelligence technology,which improves the intelligent ability of artificial intelligence and greatly promotes the development of human intelligence.It can be extended to some decision-making problems that were difficult to deal with in the past.Among them,the Multi-Agent Reinforcement Learning Algorithm(MADDPG)in the deep reinforcement learning algorithm(DRL)family has outstanding performance,with high non correlation,strong self-adaptive and self-learning functions,which can provide some efficient solutions for the perception decision-making of complex systems.This paper focuses on the theoretical application and empirical test of maddpg in A-share.First of all,this paper reviews the previous theories of deep learning,reinforcement learning,deep reinforcement learning and Multi-Agent Reinforcement Learning as well as the related academic research theories and the latest related academic literature materials in the field of Finance and investment,summarizes and comments on some previous theoretical research and application results of deep reinforcement learning in China's financial market,and focuses on clarifying The research direction and problems need to be solved.Secondly,based on the theoretical models of deep learning,reinforcement learning and deep reinforcement learning in the field of artificial intelligence,this paper constructs a model of Multi-Agent Reinforcement Learning Algorithm(MADDPG)for timing and stock selection in A-share market,and analyzes the timing and stock selection strategy through the framework of the above-mentioned stock selection strategy model.Finally,in the empirical part,the investment performance of Multi-Agent Reinforcement Learning timing and stock selection strategy is compared with the average investment performance of the selected stocks and the performance of Shanghai Stock Exchange 50 index,and the performance indicators of alpha,sharp rate and other stocks are obtained,which verifies the effectiveness of the application of the above model.The results show that the Multi-Agent Reinforcement Learning timing and stock selection model has strong data processing ability,can better deal with the big data in the financial market,can improve the ability to extract the characteristic information of transaction signals in the aspect of timing and stock selection,can also obtain some excess returns,and has strong robustness.