Scaling up the microeconomic dynamic discrete choice model of activity-based scheduling

• In this paper, the author develops a dynamic microeconomic discrete choice model in order to model mode choice and departure time in an activity-based framework. In a dynamic model, the order of activity matters and the value of time will be dependent on clock-time, preferences, previous choices and future opportunities during the day. However, the curse of dimensionality is a real problem when dealing with real-sized problems. In this paper the author argues that the effective dimension often can be reduced. Earlier evidence suggest that the effective dimension may be quite manageable. The author uses the Restricted Boltzmann Machine (RBM) to explore the dimensionality of reduction techniques. After having tuned the learning parameters of this model, the authors show that a one-layer RBM is able to find good “activity skeletons”, i.e. reasonable activity patterns that match the data. These one-layer skeletons can then be used to train a two-layer RBM which is fined-tuned to fit the data even better. The results demonstrate that the dimension can be brought down such that the model can be solved for real-sized problems.

Subject headings

TEKNIK OCH TEKNOLOGIER  -- Samhällsbyggnadsteknik -- Transportteknik och logistik (hsv//swe)

ENGINEERING AND TECHNOLOGY  -- Civil Engineering -- Transport Systems and Logistics (hsv//eng)

Publication and Content Type

vet (subject category)

kon (subject category)