Research And Application Of Financial Time Series Based On GARCH And Neural Network Model

Stocks are the main component of the financial market,and the market is changing rapidly.Traditional technical analysis methods are difficult to accurately predict changes in the stock market.A large number of researchers use machine learning to research and analyze stock time series.How to extract highly relevant stock data features and build efficient machine learning models has become the focus of current research.The mainwork of this paper is as follows:(1)Based on the GARCH model,two random disturbances and conditional variances of data features are proposed,which are highly correlated with stock volatility.First,the obtained stock yield data is verified for stationarity and normality,a GARCH model is constructed,and the features in the model are extracted to verify the correlation.The extracted stock features were tested with basic stock features and common stock data technical indicators,and it was found that new features significantly improved the accuracy of data volatility prediction.Combining new data features,technical indicators,and stock trading volume produces highly relevant data features.Compared with related papers,the random disturbance term ut is a new feature proposed in this paper based on the GARCH model,and the data feature in this paper significantly reduces the error.(2)Propose a stock prediction model based on CNN-D-LSTM model.The paper proposes a model that combines two neural networks,LSTM and CNN,in view of the relatively single type of stock prediction models.In the related research based on the combination of LSTM and CNN,this paper proposes the CNN-D-LSTM model.First,the data feature distribution is input into the LSTM,CNN network,and the LSTM is used to extract the temporal features,and the CNN extracts the static features;The output features of the LSTM are input into the LSTM to extract the time-series features of the static features;then the output features of the two LSTM layers are fully connected and the prediction results are finally output.A comparative experiment was conducted to compare the prediction error,the accuracy of trend prediction,and stock returns,and improve the accuracy of stock volatility prediction.(3)Based on the above research,a stock volatility prediction system was designed and developed.The data flow and functional design of the stock forecast are expounded,which is beneficial to stockholders to obtain more scientific and comprehensive stock data.