Enhancing Purpose Segmentation in Operational Transport Models. The Case of IMPACT, RATP’s Disaggregate Model for Paris Area

Enhancing Purpose Segmentation in Operational Transport Models. The Case of IMPACT, RATP’s Disaggregate Model for Paris Area




In this paper we discuss the need of having an adapted trip purpose segmentation in operational, full-day choice models. We focus on RATP’s (main Public Transport operator in the Paris area) disaggregate model, IMPACT (from 9 to 16 trip purposes).


Transport choice models are usually used to simulate individuals’ response to a change in a transport system (new transport supply, transport policies, socioeconomic evolutions, etc.). In order to do so, they need to oversimplify the transport system as well as individuals’ mechanisms of choice to try to reconstitute them into a transport model. Indeed, when residents of an urban area define their activity programs and the trips they need to do, they take into account a large number of factors, which can be related to individual and household characteristics (e.g., gender, age, income or number of cars), transport supply (e.g., level of service or costs), spatial characteristics (e.g., number of jobs or walkability) and activity characteristics (e.g., time-of-day or purpose).
In this paper we discuss how trip purpose is taken into account in transport choice models, and the necessity of having an adapted purpose segmentation. Indeed, some operational choice models consider that all individuals choose their transport mode or their trip destination irrespectively of trips purpose. Others only focus on purposes related to peak hours (i.e. home-work, sometimes home-education). But, if we want to simulate mode and destination choices on a full-day basis, we need to work with all trip purposes and differentiate them.
This has been the approach of the RATP’s (main Public Transport operator in the Paris area) disaggregate choice model, IMPACT. For more than twenty years, IMPACT segmented trips into nine different categories for modelling: home-work, home-education, home-employer’s business, home-shopping, home-leisure, home-personal business, home-escort, non-home based employer’s business and other non-home based trips.
But, the last household travel survey in the Paris area, the Enquête Globale de Transports from 2010, more detailed than the previous one from 2001, allows us to segment even more accurately trip purposes. Indeed, in 2010 almost 125,000 trips have been described with 38 different purposes, when in 2001 we had only 80,000 trips and 20 purposes.
This more accurate representation of trips purposes let us improve IMPACT’s capacity of simulating mode and destination choices. By passing from 9 trip purposes to 16, we can distinguish individuals’ attitude towards different kinds of trips which were treated equally in the previous versions of IMPACT. For instance, home-shopping trips have been split in three different purposes (home-daily shopping, home-weekly shopping and home-exceptional shopping). This way, we can distinguish short and iterative trips made commonly on foot or on the way home at local retail shops from once-a-week trips, made typically at malls by car, or from more occasional trips, made in specialized shops, located in areas with or without an efficient public transport supply.
In this paper we analyze household travel data to show the heterogeneity of individual trip choices when they travel for different purposes. Then, we discuss the segmentation process and the criteria we used when we passed from 9 to 16 trip purposes. Finally, we present the model estimation results which confirm the need for a more detailed purpose segmentation.


Association for European Transport