Linear addition is commonly used to generate ensemble forecasts for decomposition ensemble models but traditionally treats individual modes with equal weights for simplicity. Using Taiwan air passenger flow as an empirical case, this study examines whether incorporating weighting for individual single-mode forecasts assessed by grey relational analysis into linear addition can improve the accuracy of the decomposition ensemble models used to forecast air passenger demand.
Data series are decomposed into several single modes by empirical mode decomposition, and then different artificial intelligence methods are applied to individually forecast these decomposed modes. By incorporating the correlation between each forecasted mode series and the original time series into linear addition for ensemble learning, a genetic algorithm is applied to optimally synthesize individual single-mode forecasts to obtain the ensemble forecasts.
The empirical results in terms of level and directional forecasting accuracy showed that the proposed decomposition ensemble models with linear addition using grey relational analysis improved the forecasting accuracy of air passenger demand for different forecasting horizons.
Accurately forecasting air passenger demand is beneficial for both policymakers and practitioners in the aviation industry when making operational plans.
In light of the significance of improving the accuracy of decomposition ensemble models for forecasting air passenger demand, this research contributes to the development of a weighting scheme using grey relational analysis to generate ensemble forecasts.
