Spatio-temporal autoregressive models of monthly air passenger flows

Abstract

In nature, origin-destination air passenger flows are correlated both spatially and temporally due to spatial and temporal relationships of human behaviors and environments. However, most existing studies for modeling air passenger flows have assumed these relationships to be independent

few studies have considered either spatial or temporal dependences. To consider both, we develop spatio-temporal autoregressive models of monthly origin-destination air passenger flows. Benefited from the special structure of a spatio-temporal dependence matrix in our model, the proposed model incorporates multidimensional autoregressive coefficients that enable more effective predictive performance. Its application to real open-access aviation data in 2010 demonstrates the effectiveness of the proposed models in forecasting monthly air passenger flows.