NASDAQ INDEX PREDICTION BASED ON ARIMA-GARCH MODEL AND DYNAMIC REGRESSION

As the global market becomes more and more open and volatile, it is of great significance to grasp the financial temporal volatility and correlation and accurately predict stock price behavior. This paper takes the Nasdaq Composite Index in 2020-2023 as the research object, constructs the ARIMA(1,1,(1,5)) model with GARCH(1,1) conforming to t distribution disturbance term to fit the trend of the index in 2020-2022, and predicts the trend in the first half of 2023. The results show that the model with GARCH effect gives a wider forecast confidence interval and can indicate the potential risk, but can not accurately reflect the real trend of the index. To improve the prediction accuracy, this paper takes Nasdaq index as the response variable, introduces S&P 500 index as the input variable, constructs an effective dynamic regression model through Granger causality test and EG cointegration test, and improves the model through cross-correlation function analysis. The results show that the forecast trend of the model is closer to the actual series of fluctuations, indicating that the S&P 500 index plays a promoting role in predicting the Nasdaq index, which provides a more reliable reference for investors when weighing the benefits and risks.