Research And Implementation Of Auxiliary Securities Investment Method Based On Deep Learning

In recent years,with the continuous development of the security market,the number of Chinese investors has reached a new high.People hope to realize the wealth preservation or asset appreciation in the process of security investment.How to dynamically adjust the assets of all underlying securities so as to reduce investment risks and increase investment returns is an important research problem in the field of securities investment.However,many factors in the market will affect the rise and fall of securities prices,which makes the securities prices have great volatility and are difficult to predict.Secondly,most investors in our country lack professional financial knowledge,so it is difficult to analyze useful information from the market data.Finally,portfolio strategies also face the challenge of high reallocation costs or even failure due to the inability to adapt to complex market fluctuations.In order to solve the above problems and challenges in the process of securities investment,this thesis studies the method of investment portfolio,and designs and implements a securities investment auxiliary system to help investors analyze the current market conditions,and give risk tips and operational suggestions on the underlying stocks.The main work of this thesis is as follows:(1)A multi-agent cooperation stock investment framework MACSIF based on MADDPG algorithm is proposed.Firstly,this thesis designs different reward functions to guide agents to give robust,neutral and aggressive investment suggestions respectively,and then integrates the investment decisions of multiple agents to form portfolio suggestions.Secondly,aiming at how to deal with the complex and changeable market environment,effectively obtain effective information from the complex market and give investment suggestions,this thesis designs a multi-agent hierarchical feature extraction encoder to obtain transaction characteristics,corporate financial data characteristics,corporate market characteristics and global market characteristics respectively.(2)In view of the high replacement cost of portfolio strategy and the current situation of buying and selling restrictions in our securities market,this thesis proposes to add the mechanism of buying and selling maximum and minimum quota restrictions into MACSIF’s portfolio strategy,so as to adapt to the real investment scenario.Further,in order to advance the future found earlier downside risk,this thesis designed a used for investment advice to the risk of the structure of the secondary check,by a more accurate the short-term stock price fall risk prediction model,predict the future 5 days of share price change,and adjust the portfolio according to the results of the prediction are suggested.(3)In order to provide investors with necessary information to help them analyze the current market conditions,this thesis collects a large number of securities related data through crawler,and designs and implements an intelligent investment assistance system.The system provides basic information of listed companies and visual financial statement data to help investors quickly understand the investment target.In addition,it also provides the function of user-defined stock selection strategy and backtest,allowing users to choose their own investment strategy.The above stock price drop risk prediction model and MACSIF are deployed on this system to give risk tips and operation suggestions for investors respectively.