On Bicycle Choice Set Generation

Abstract

In this thesis, we studied various methods of choice set generation. The Double Stochastic Generation Function (DSGF) method was looked after in more detail. We showed that the routes provided by this method are too complex in terms of the number of Basic Path Components (BPC). This problem is very clear when the predicted routes are compared with the observed routes provided by the FietsTelweek. In various ways we tried to cope with this problem. First we tried to scale back the cumulative distribution function (cdf) of the predicted data to the cdf of the observed data. Secondly, a maximum likelihood function was introduced to keep the most interesting predicted routes in terms of BPC's. The latter method proved to be more useful in constructing a choice set containing reasonable, not too complex routes.