Comparison of Parametric and Nonparametric Hazard Models for Stop Durations on Urban Tours with Commercial Vehicles

Hazard-based stop duration models for the stop duration of commercial vehicles in urban areas are presented. Passively collected GPS data were used to estimate two hazard-survival models to predict the stop duration of commercial vehicles undertaking urban pickup and delivery tours. The first was an accelerated failure-time parametric hazard model, and the second was a proportional nonparametric hazard model. Explanatory variables included time of day, population or employment density, number of stops, distance of inbound and outbound trips, and attributes of destination establishments, such as sales volume and industry classification. Models were estimated with and without establishment data because these data may not always be available for model application. Results showed that passively collected GPS data could be used to estimate stop duration models when linked with other data sources that provided appropriate explanatory variables. The parametric models were shown to outperform the nonparametric models and had higher measures of goodness of fit and better hazard distributions for stop duration.