Research on Stock Price Volatility Prediction based on Deep Neural Network

Traditional volatility models are challenging to deal with nonlinear relationships in time series and are only suitable for time series prediction with small-scale single variables, which leads to low accuracy of volatility prediction. To solve the above problems, this paper proposes a new method of predicting financial market volatility based on deep learning. This method first selects the features of financial factors using the counterfactual method, then aims at the nonlinear and long-term dependence problems of volatility. A volatility prediction model NN-FinTrans (Neural Network Finance Transformer) based on an improved Transformer is constructed, and experiments are carried out on the data to verify the effect of the model prediction.