Application Of CNN And LSTM In Short-term Stock Price Rise And Fall Prediction Of Cyclical Stocks

The stock market is a dynamic complex nonlinear system.Its future trend has always been considered difficult to predict,but accurately grasping the future trend of the stock market has important practical significance for regulators,investment institutions and individuals.With the continuous development and improvement of mathematical statistical methods,using mathematical models to accurately quantify the future trend of the stock market is still a hot topic.In the process of research advancement,many methods of linear model and nonlinear model have emerged.Among them,machine learning model has gradually attracted the attention of researchers because of its strong ability to represent nonlinear data,and has achieved good results in stock market prediction.Combined with the rapid development of modern digital finance and the expansion of data volume,using machine learning model to study the high-frequency data of stock market has become one of the hotspots of global financial market research.Compared with low-frequency data,high-frequency data has higher timeliness and accuracy,and plays an important role in short-term trend prediction.Based on the above background,this paper selects all the industry constituent stocks of coal and steel(Shenwan industry standard)in the strong cycle industry of a shares as the research object,and uses the convolution neural network model(CNN model)and long and short-term memory model(LSTM model)in deep learning,which are characterized by the influencing factors in the prediction of stock future trend,To explore the application of deep learning in the field of stock high-frequency data.CNN model is mainly used to classify the trading patterns of strong cycle stocks in different time periods,and distinguish different trading periods and states of strong cycle stocks as far as possible,so as to distinguish different distributions;LSTM model is responsible for predicting the rise and fall of stock price and evaluating the effect of the model based on the data of the same trading mode.At the same time,the parameter configuration of the model is continuously optimized in the training process to find the parameters that make the model give full play to the optimal effect.In the first mock exam of the above two deep learning models,this paper conducts an empirical analysis from the perspective of single model and combined model.Experiments are carried out on the linear model,CNN model,LSTM model and CNN and LSTM mixed models.The prediction accuracy of the prediction is based on the future 1min stock price rise and fall.Through empirical research,it is found that the combined model of CNN + LSTM has better prediction accuracy than the single linear model;It shows that strong cyclical stocks do have different trading modes in different periods,and better results can be obtained by classifying them and predicting them separately.This method can provide some reference significance for future quantitative institutional workers to study the corresponding strategies.At the same time,it also shows that starting from economic theory and combined with the advantages of deep learning,it will have broader application prospects,and the development of digital finance can be expected in the future.