Hong Kong Air Traffic: Explanation and Prediction based on Sparse Seasonal ARIMA Model

Abstract: The monthly air traffic of a city is a time series with an obvious seasonal pattern, and is closely related to the economic situation and social environment of the city. In Hong Kong, for example, July, August, and October tend to be the peak season of traffic flow, while there is also a relatively fixed off-season. In the case of a stable social environment, a carefully identified and fitted seasonal ARIMA model can predict the traffic flow in the future months well. This work selects the air traffic data, including arrival and departure passengers of Hong Kong, after the financial crisis and before the political storm. A sparse seasonal ARIMA$(0,1,1)\times(4,1,0)_{12}$ is built, which can correctly predict the air traffic from January to July in 2020 within its $95\%$ confidence interval. Furthermore, this work decomposes the time-series and find that important events, like the financial crisis, political storm, and the COVID-19 outbreak, affect the level of air traffic to some extent. For example, the political storm and epidemic prevention and control that happened after 2019 made the air traffic drop significantly. According to my sparse seasonal ARIMA model, the air traffic from February to November in 2020 is only $5\%$ of what it should be without these two events. This is a valuable application of the time-series model in the air traffic loss estimation.