Prediction Of Stock Price Model Based On NPCA-MM-LSTM

With the innovative development of artificial intelligence technology,the financial market produces a large number of financial information with sequential,temporal and non-linear correlation.How to reasonably process and analyze these data and build the stock trend prediction model is a very challenging and important academic work.To solve this problem,this paper applies machine learning and deep learning algorithms to the field of stock trend prediction,and uses financial data to build a scientific and effective financial time series prediction model in line with market demand.The main research contents of this paper are as follows:(1)For financial data combined with data preprocessing technology,data normalization is used to plan the share price characteristics for the same order of magnitude of data.MPCA was used to reduce the high-dimensional data of stocks using MPCA and NPCA.Stock data were matrix-decomposed using SVD,using MDS to retain similar information about the maximum separability and high-dimensional space of the data.(2)Using RNN to predict stock trend fluctuations: RNN can effectively deal with financial time series problems,and then avoid over-fitting phenomenon.The NPCA-RNN model was used to predict the stock trend,and it is verified that the accuracy of the model on the stock price fluctuation is better than that of the NPCA-SVM model,which is improved by about8%.(3)The NPCA-MM-LSTM prediction model based on the analysis of financial time series information is proposed.This model adopts the 13-dimensional data of Vanke A-shares,uses NPCA method to obtain 3-dimensional main components with cumulative contribution rate higher than 85% from 13 financial characteristics,and input them to MM-LSTM neural network for model training,forming the NPCA-MM-LSTM model.Using KDJ as the quantitative investment standard,the yield of the model is higher than 10% of the stocks in the actual quantitative transactions,thus enriching the application of deep learning method in the financial field and providing effective and reasonable market investment suggestions for investors.