A0520
Title: A hybrid smoothing approach for estimating airline passenger demand
Authors: Maria Rosa Nieto Delfin - Investigaciones y Estudios Superiores, S.C (Mexico) [presenting]
Rafael Bernardo Carmona Benitez - Investigaciones y Estudios Superiores, S. C. (Mexico)
Abstract: The focus is on a hybrid model for estimating air passenger (pax) time series at the route level, thus addressing the limitations of existing unobserved components models, which primarily estimate trends. Accurate modeling of passenger demand is crucial for the air transportation industry, where time series exhibit strong seasonal behavior influenced by socioeconomic conditions and exceptional events. A methodology is proposed that integrates the trend-based unobserved components model developed by a prior study with the Holt-Winters multiplicative model. This approach enables the estimation of trend, seasonality, and cycle components in a unified framework. The proposed model involves the multiplication of the estimated trend component by the seasonal and cyclical components. This process enables the reconstruction of the original time series with greater precision. The hybrid model is validated using U.S. Bureau of Transportation Statistics route-level data, and its estimating accuracy is compared against the classical decomposition method and the Holt-Winters model. The performance of the model is assessed using the Diebold-Mariano test and mean absolute percentage error (MAPE). The findings indicate that the integration of a hybrid model enhances estimation accuracy by capturing the multi-component nature of pax time series. This approach provides a flexible and interpretable framework for time series in sectors with strong seasonal dynamics.
