Monte Carlo's Simulation to Measure the Uncertainty of Travel Demand and Examining the Impact of Quasi-Random Random Number Generators

Lack of uncertainty studies is one of the common shortcomings in travel demand forecasts. This has led to overestimation of demand and has made some transportation infrastructures unprofitable. Investing in projects with less but more reliable profit is important. Mount Carlo simulation is one of the most common methods for uncertainty and sensitivity analysis. One of the main requirements of Monte Carlo is the use of efficient random number generators to generate random numbers with high uniformity. But the uniformity of random numbers produced by pseudo-random number generators (PRNG’s) may be sometimes good and sometimes poor. But quasi-random number generators (QRNG’s) produce a sequence of deterministic random numbers with more uniformity and thus better filling the unit hypercube. Therefore, it is necessary to examine and quantify the uncertainty of travel demand models and investigate the effects of RNG’s. The aims of this research article are threefold: 1. Measure the uncertainty of travel demand (production and attraction) models; 2- Using sensitivity analysis, rank the input variables that play the most important role in the uncertainty of the model outputs; and 3- Investigate the effect of quasi-random and pseudo-random generators on uncertainty. The results of this study showed that the travel production and attraction models reduce the uncertainty of inputs and the uncertainty of the attraction models is higher than the production models. The Latin hypercube sampling (LHS) and shuffled Halton are marginally more stable than other methods assessed in this study.