Stock investment contains many potential risks because of the volatility of stock price. Therefore, forecasting the stock index has become one of the most favorite research topics of investors and researchers all over the world. This paper aims to introduce the hybrid ARIMA-GARCH model, which is commonly used beside the single ARIMA and GARCH models for forecasting the time series; and apply it to predict the VN-Index in Vietnam's stock market. The comparison among the three models shows that the hybrid ARIMA(1,1,1)-GARCH (1,1) model provides better predictions than the two single models including ARIMA (1,1,1) and GARCH (1,1).