The stock market is a barometer of the real economy,and the forecast of the rise and fall of stock prices has important economic and social value.The rise and fall of stock prices is affected by many factors,and there is a memory relationship between stock time series data.Long Short Term Memory can process time series data types with long-term dependence and mine hidden data in nonlinear data.information.Therefore,this paper builds a stock price trend prediction model based on long-term and short-term memory neural networks.The research content of this paper is mainly divided into four aspects.First,when the research object is selected,by comparing the size of the company,the company's industry status and the company's financial status,the company's stock is selected as the research object.In the process of selecting and processing the feature data,in addition to the technical indicators,transaction data and financial indicators,the net transaction volume data of the day-to-day large singles is added.In order to reduce the impact of stock price fluctuations on the forecast results during the stock cycle,data preprocessing selects the smooth normalization method.Second,due to the multi-dimensionality of the input data and the complex relationship between the dimensions,Principal Components are selected in this paper.Analysis and Denoising Autoencoder feature extraction methods are used to extract the input data.By comparing the number of dimensions and Mean Square Error of the two feature extraction methods,the DAE effect is better than that of the two.PCA;Third,in the LSTM model hyperparameter setting,the comparative analysis of time step,hidden layer neurons,learning rate,etc.,finally determine the LSTM model structure;fourth,in the analysis of experimental results The prediction accuracy of the three models of LSTM model,PCA-LSTM model and DAE-LSTM model on the stock price trend of China Ping An Company.The model is used to predict and analyze several other weighted stocks.The final result shows that the LSTM model is suitable for weighted stock trend forecasting.The prediction accuracy of PCA-LSTM model and DAE-LSTM model is much higher than that of LSTM model and DAE-LSTM model.The effect is better than the PCA-LSTM model.The study is composed of four parts,data processing and feature extraction in the research of domestic stock price trend forecasting.It has certain practical significance to predict the price trend of individual stocks and then assist in trading decisions. |