ABSTRACT
This chapter deals with a forecasting framework based on the seasonal decomposition method with the novel use of Google Trends data as the exogenous variables to capture the moving holiday effect. It compares the out-sample forecasting results for January and February: with Chinese New Year (CNY) regressors generated through the Genhol program; with Google Trends data as exogenous variables quantifying the CNY effect; and with no CNY factors. In literature of the short-term Chinese air travel demand forecasting, people can found that the forecasting errors for Januarys and Februarys are usually larger than for other months. The chapter presents a nowcasting process in detail with weekly Google Trends data by adopting Mixed Data Sampling Regression models, to show that the information with a higher frequency can effectively help to improve the forecasting accuracy. It provides an empirical analysis and comparison between the common approach and the proposed method and describes a forecasting framework based on the former analysis.
