Research on Stock Price Volatility Prediction Based on Generative Adversarial Network

Abstract

In order to explore the application effect of the most popular Generative Adversarial Network (GAN) in the field of financial forecasting, this paper proposes to explore the predictive ability of GAN's stock price volatility by taking the daily closing price of the S&P 500 index as the research object. The empirical method takes EGARCH model and Long Short-Term Memory (LSTM) as the benchmark model, MSE and MAE as the prediction error measurement indicators, and empirically compares the prediction results of the three models to analyze the out of sample prediction ability of GAN one day in advance. The empirical results show that GAN has the lowest prediction error and the highest prediction accuracy. LSTM also has a good prediction effect, but it is slightly inferior to GAN. EGARCH model has the largest prediction error. It shows that GAN, as a cutting-edge deep learning technology, has a good application prospect in the field of financial time series prediction.