Study On Stock Price Trend Prediction Method Based On Deep Learning

The method of stock trading strategy and stock price trend forecast is a hot topic in recent years,but at present.Due to the lag of stock trading signals and the low accuracy of trend prediction that the judgment of stock trading signals faces many challenges.In this paper,The method of studying stock trading strategies is based on indicators such as DMI,KDJ and MACD,Research on stock price movement methods,based on regression methods and deep learning methods.The main research contents are as follows:?1?On the basis of deep learning and multiple regression methods,the multiple regression based on deep learning model is established,which will be applied to predict the forecast of stock price movements.Based on the DenseNet network,the ReLU activation function is used between the layers,and the MRDL₃ and MRDL₄ models of the stock opening price,the highest price and the lowest price are input variables,and the closing price is the output variable.In order to avoid the local minimum when training the regression coefficient.The MRDL₃ and MRDL₄ models use the MBGD optimization algorithm to optimize the regression coefficients.?2?Based on the DMI indicator,the long-short indicator DMISV trading strategy is proposed.This trading strategy determines the rate of stock price growth or downtrend by calculating the DDI index.Based on the K value in the KDJ indicator and its magnitude of change?DK?,the KDJSV trading strategy method is presented which utilzing the changes of overbought and oversold.The MACDV trading strategy based on the MACD indicator is demonstrated which determines the changes of the strength of both the multiparty and empty parties in the stock market.Finally,a DKB trading strategy integrated DMI,KDJ,and MACD indicators is proposed.The experimental data was downloaded from the website of Juchao Information,including the daily data of the Shanghai and Shenzhen stock markets from October 1,2013 to October 17,2018.It is used to verify the effectiveness of the above method.The experimental results show that the MRDL used to predict the stock closing price trend,and the mean square error is between [0.0043,0.0821].The DMISV,KDJSV,and MACDV stock trading strategies presented in this paper 80% of the stock that the return rate are more than 10%.The experimental results verify the effectiveness of the proposed trading strategy and stock price trend forecasting method.