Does the optimal model always perform the best? a combined approach for interval forecasting

Interval forecasting is widely applied by decision makers for it can provide more comprehensive information. In the literature, GARCH models under different distributional assumptions are applied and evaluated to find the optimal interval forecasting model for the experimental data. However, the optimal model selected based on sample data from a specific period may not always perform the best in future periods. Therefore, this study employs GARCH models based on different distributional assumptions for interval forecasting of the daily return data of the Nasdaq Composite Index. The results show that the forecasting performance of some models exhibits significant differences across different periods. To address this issue, this study proposes a Monte Carlo-based non-parametric interval forecasting combination method. The results demonstrate that this method can effectively avoid the risk of forecasting inaccuracies caused by relying on a single model.

Actions (login required)

View Item