Abstract Manager

Disaggregate travel demand models in practice involve random numbers to convert probabilistic outcomes of the logit choice models into discrete individual choices. It is important to evaluate the model properties with respect to the microsimulation variability and ensure that the model outcomes are possible to verify on the disaggregate level. To provide a valid assessment of the projects and policies the demand impacts have to be stable and independent of the random numbers used in the microsimulation.     

The analysis and comparisons include three different principal methods for the conversion of probabilistic model outcomes to discrete individual choices:

• Monte-Carlo with Uncontrolled Random Numbers (MCR).
• Monte-Carlo with a fixed seed for each individual choice (MCF).
• Random utility simulation with fixed seed (RUF) for each individual and choice alternative that is generated in advance and kept fixed across compared scenarios.

These methods can be evaluated at two levels – aggregate and disaggregate, with respect to the following properties and metrics:

• Repeatability of the results with the same inputs.
• Continuity across the range of inputs from large differences to small differences.
• Monotonicity that means logical elasticities across the range of scenario inputs.   
• Comparability that requires a stable difference between two scenarios across different random seeds (Although MCF and RUF methods inherently use fixed random seeds across compared scenarios, the choice of the original seed remains arbitrary).

The insights are illustrated first for a single choice dimension (one sub-model), and then for a sequence of interlinked sub-models in the Regional Travel Model for Edmonton modelling region. It is shown that only RUF can ensure all required properties (at both aggregate and disaggregate level) while MCF has only some of them (primarily at the aggregate level), followed by MCR as the most volatile method that may violate any of the desired properties. The differences between the three methods at the aggregate level might be minimal when the scale of the simulation (number of affected individuals) and magnitude of differences between scenarios are large. The impact of the choice method is analyzed using such projects as bike lane improvements, bridge closure, new transit mode, etc. The results show that RUF ensures logical elasticity and substantially reduces so-called microsimulation noise in comparing the impact of the projects.