Abstract:

This study compares the forecasting accuracy and model fit of two common time series models, the Autoregressive Integrated Moving Average (ARIMA) and the Generalised Autoregressive Conditional Heteroskedasticity (GARCH), for financial data. Performance measures included MSE, RMSE, MAPE, R-square, AIC, and BIC. The GARCH model outperforms the ARIMA model in various ways. one experiment, the GARCH model performed better at forecasting financial time series, with lower MSEs and RMSEs. According to MAPE, the forecasts based on GARCH were more accurate than those based on ARIMA. Although AIC and BIC have negative R-squared values, GARCH models performed better. It indicates that the data variance assumptions were violated. This study emphasizes model selection for time-series forecasting, especially in financial markets, where precise predictions are crucial. The improved fit of the GARCH model suggests it can model volatility in financial time series.