Gravity model for forecasting airline passenger flow considering network structure

In order to improve the precision of airline passenger flow prediction, the effects of spatial distance between cities, degree, average path length, k-core, edge betweenness and point connectivity on airline passenger flow are explored. And then, a gravity model for forecasting airline passenger flow is constructed. Prediction accuracy of the gravity model proposed in this study reaches 0.87, so the goodness of fit is acceptable. The results indicate that the influences of spatial distance on airline passenger flow have a characteristic of duality. The greater the degrees of two cities are, the smaller the proportions of passenger flows corresponding to each airline are, and the greater the traffic impedance between them is. Edge betweenness is not the key factor affecting airline passenger flow. Moreover, Average path length, k-core and point connectivity significantly affect airline passenger flow, but there is multicollinearity between these indicators and spatial distance (or degree).