Route sampling correction for stochastic route choice set generation

Bovy, Piet H. L., Bekhor, Shlomo and Prato, Carlo Giacomo (2009). Route sampling correction for stochastic route choice set generation. In: Transportation Research Board (TRB) 88th Annual Meeting. Transportation Research Board (TRB) 88th Annual Meeting, Washington, DC, United States, (). 11-15 January 2009.

Author Bovy, Piet H. L.
Bekhor, Shlomo
Prato, Carlo Giacomo
Title of paper Route sampling correction for stochastic route choice set generation
Conference name Transportation Research Board (TRB) 88th Annual Meeting
Conference location Washington, DC, United States
Conference dates 11-15 January 2009
Proceedings title Transportation Research Board (TRB) 88th Annual Meeting
Publication Year 2009
Sub-type Fully published paper
Total pages 18
Language eng
Abstract/Summary Contexts involving a very large number of choices are traditionally modeled by estimating Multinomial Logit (MNL) models after sampling from the full set of available alternatives. Route choice in typical urban networks may have a large number of alternatives, but MNL is not suitable to model route choice. In addition, the sampling approach introduced by stochastic route-set generation procedures does not satisfy the property of equal selection probability of the choice set generated. Stochastic route set generation is a case of sampling importance because the selection probability of a route depends on the properties of the route itself. This paper presents the calculation of the selection probabilities and generation frequencies of routes when this sampling process is implemented and used for the estimation of route choice models. A correction is shown to be needed only in cases of samples with unequal route selection probability. Both the large sample case and the equal selection probability case can be treated the same, namely without correction for sampling. This also holds for enumeration methods of choice set generation. The results of an experimental route sampling correction are presented for two real datasets. The experiments presented validate the possibility of using unequally sampled choice sets in estimating choice models, while supporting the use of stochastic route generation methods.
Q-Index Code E1
Q-Index Status Provisional Code
Institutional Status Non-UQ
Additional Notes Transportation Research Board Annual Meeting 2009 Paper #09-0306

Document type: Conference Paper
Collection: School of Civil Engineering Publications
 
Versions
Version Filter Type
Citation counts: Google Scholar Search Google Scholar
Created: Tue, 19 Apr 2016, 16:16:31 EST by Carlo Prato on behalf of School of Civil Engineering