Multilevel modelling of commercial vehicle inter-arrival duration using GPS data

• We develop models of inter-arrival duration between carrier visits to destinations.
• The models are based on longitudinal GPS data linked to other secondary data sources.
• Multi-level multinomial logit models outperform multilevel ordered probit models.
• Market segmentation improves model results.
• Many carrier firms do not follow consistent shipping schedules.

This study uses truck GPS data to study inter-arrival duration, defined as the time between arrivals at a destination of two successive vehicles operated by the same carrier. Destinations were separated into market segments: (1) frequently-visited destinations, (2) regularly-scheduled destinations, and (3) unscheduled destinations, based on visit frequency and regularity. Inter-arrival duration was modelled using multilevel ordered probit and multilevel multinomial logit models. Market segmentation improved modelling results, and multilevel models performed better than single level models. Results showed a wide variation in shipping behavior of commercial establishments and that many firms do not follow consistent shipping schedules.