Research On Single Stock Trading Strategy Based On Investor’s Comprehensive Sentiment And Reinforcement Learning

In stock market,trading decisions made by individuals are not only influenced by market data but also by the emotions and attitudes of others.Lacking professional knowledge regarding to stock trading,most investors cannot make rational judgments on the information they receive,resulting in significant losses.Therefore,investors capable of analyzing stock data and making rational trading decisions are exceedingly important in the healthy development of the stock market.This article proposes a single stock trading strategy based on comprehensive investor sentiment and reinforcement learning algorithms to reduce trading risks.The conclusions can be made from the following three aspects:Firstly,an investor’s comprehensive sentiment index was constructed.Based on the influence of individual and institutional investors,600,000 stock forum comments representing individual investor opinions and 2,000 institutional research reports representing institutional investor attitudes were collected.Using Vader sentiment analysis to analyze the sentiment tendency of the text and constructing an investor’s comprehensive sentiment index to measure the market sentiment of the stock.Secondly,a single stock trading strategy based on investors’ comprehensive sentiment and DQN was studied.In terms of stock market data and investor comprehensive sentiment indicators,the deep reinforcement learning DQN algorithm was used to simulate stock trading,and experiments were conducted on four stocks.The results showed that investors’ comprehensive sentiment helps to improve the stability of investment strategies and reduce trading risks.Thirdly,a single stock trading strategy based on investors’ comprehensive sentiment and DRQN was explored.In view of the drawback that a fully connected network cannot acquire historical experience,the LSTM network was used to replace the fully connected network,and the moving average strategy was combined to reduce trading frequency.Experimental verification showed that using DRQN to simulate stock trading resulted in more robust returns.The single stock trading strategy proposed in this article can effectively reduce trading risk in the securities trading process.Applying sentiment analysis and deep reinforcement learning to the securities market can help investors understand market trends better and predict future market trends more correct,and eventually make wiser investment decisions.