Hierarchical forecasting based on AR-GARCH model in a coherent structure

This paper compares the accuracy of the aggregate forecasting with the bottom-up forecasting based on AR-GARCH model for the return rate of simulated Dow Jones Industrial Average. Most of the existing stock price index studies did not consider the hierarchical structure and often missed the coherent relationships between individual components. In this experiment, we simulated 30 coherent components based on AR(2)-GARCH(1, 1) model. Then we evaluated the performance of both forecasting methods ignoring the coherent structure. The results of our experiment indicated that the accuracy of forecasting method varied depending on the correlation degree of 30 coherent components, however the data noise did not significantly influenced the performance of hierarchical forecasting method.