PLR-IRF Turning Point Prediction And Optimization DQN Stock Quantitative Investment Decision-making Model

In recent years,with the rapid development of computer technology,quantitative investment has gradually emerged in the field of financial investment,and the quantitative investment of stocks is the most representative.However,in the stock market,stock trading is frequent and complicated,and stock price trend prediction is difficult,which makes it difficult for ordinary investors to grasp its changing rules.In addition,many quantitative investment strategies are only to realize the trading ideas directly with the help of computers,and the strategies themselves are less adaptive and less versatile.At present,the stock market still has problems such as low accuracy of stock price trend forecast,low investment yield,and high investment risk.In view of the above problems,this paper applies artificial intelligence technology to stock quantitative investment,and designs a new stock quantitative investment decision model.The model includes three modules:data preprocessing,trend prediction and investment decision.The data preprocessing module implements pruning and information summarization of stock data,and uses transaction data and indicator data to ensure the comprehensiveness of the data.The trend prediction module proposes a stock price trend turning point prediction algorithm PLR-IRF(Piecewise Linear Representation-Improved Random Forest),which can achieve turning point extraction,classification and prediction.It is an optimization of the Piecewise Linear Representation(PLR)and Random Forest(RF)algorithms,which can realize the automatic setting of threshold in PLR algorithm.And,the Genetic Algorithm(GA)is used to improve the RF algorithm.The investment decision module training Deep Q Network(DQN)model design investment strategy,and optimize the parameters of each part of the DQN model,so that the model can independently learn an investment strategy that enables investors to obtain more profits.Finally,using the model proposed in this paper,the experimental was carried out on 20 randomly selected stocks,and two sets of comparative experiments were designed to verify the validity and feasibility of the model.The experimental results show that the model can obtain higher stock price trend turning point prediction accuracy,better investment profit,stronger versatility,and can provide valuable reference for investors.