Application of passenger flow estimation based on network centrality indicators to reorganise the Shizuoka bus network

Recently, demand forecasting models based on regression analysis have been developed at the station level to cope with the limitations of conventional four-step models when predicting passenger flow in public transport (PT) networks. This study aimed to directly estimate the passenger flow distributions of the Shizuoka bus network using network properties quantified by several centrality indicators and socioeconomic variables in three conventional regression models: ordinary least squares, Poisson, and negative binomial. Two representations of the PT network, in infrastructure and service space forms, were applied to compute unweighted and weighted graphs. Hyperpath-based calculations were utilised to identify the optimal service attributes from a set of attractive lines as a link weight. The best fit of Poisson regression model suggested that passenger flows could be explained not only by the network properties but also by factors identifying the characteristics around station. The findings from this model are expected to be useful for informing strategic development policy during the evaluation and reorganisation of bus operations and routes for maximum efficiency.