Air Traffic Demand Forecasting with a Bayesian Structural Time Series Approach

Abstract

Airport planning, and therefore the development of air infrastructure, depends to a large extent on the demand forecast for the future. To plan investments in the infrastructure of an airport system and to be able to meet future needs, it is essential to predict the level and distribution of demand, both for passengers and air cargo. In the present work, a forecast was made, in the medium-long term (10 years), of the demand for passengers and air cargo, applied to a specific case study (Colombia), and where the impact on air traffic during the most severe period of the COVID-19 pandemic was taken into account. To achieve this objective, and as a methodological approach, a model of the Bayesian Structural Time Series (BSTS) type was developed, designed to work with time series data, and widely used for feature selection, time series forecasting, and the immediate inference of the causal impact. From the results obtained, two relevant aspects can be highlighted, firstly, both demand and its growth trend will recover very soon (in just a couple of years), compared to the pre-pandemic year 2019, which was analyzed in the case study. And, secondly, the model presents very acceptable MAPE values (between 1% and 7%, depending on the variable to be forecasted), which makes the BSTS method a viable alternative methodology for calculating air traffic forecasts.