Exploring travellers’ risk preferences with regard to travel time reliability on the basis of GPS trip records
Travel time reliability has attracted considerable interest in the field of route choice modelling. Knowing how individuals choose paths with uncertain travel times is fundamental to advancing our understanding of route choice behaviour and thus driving the development of route guidance systems. In general, existing navigation systems provide the shortest path on the basis of distance or travel time, even though many travellers do not intend to choose the shortest path. Several studies have shown that the probability of delay or travel time reliability is an important factor in a traveller’s route choice decision. Learning a traveller’s risk preference with regard to travel time reliability is important for designing a preferable route. Traditionally, route choice data for individual preference analysis are collected by conducting stated preference surveys. However, this approach is difficult to avoid its inherent limitation, namely a lack of honest, accurate, and bias-free reporting. To overcome these problems, the present study proposes a new data collection methodology that facilitates estimation of a traveller’s risk preference on the basis of large-scale GPS trip records. The lower and upper bounds of individual risk preference can be estimated by exhausting a series of reliable paths with different on-time arrival probabilities and using the theory of stochastic dominance. Then, a regression model based on a logistic function is established to explore how socio-demographic and trip characteristics influence the lower and upper bounds. Thus, individual properties, such as age, and pre-trip information, such origindestination (OD) distance, departure time, and day of week, are found to have a significant influence on the degree of risk preference.