A simulated annealing algorithm for the creation of synthetic population in activity-based travel demand model

Synthetic population serves as an agent group of the population of activity patterns that possesses identical socio-economic characteristics applied on the model that forecasts the travel demand in the activity-based profile. During the development of agentbased traffic simulation model, the accuracy of simulation outcomes relies highly on population synthesis. This study endeavors to develop a synthetic population, based on the Simulated Annealing (SA) algorithm for the activity-based travel demand model. Hill climbing and cooling schedule are essential elements to be considered when applying SA into the synthetic population. Also, Metropolis-Hasting Algorithm was employed to decide whether to select or dismiss the follow-up distribution so that hill climbing phenomenon can be prevented. Additionally, the stability of the algorithm was figured through a scenario analysis of the optimal combination of iteration and temperature T on cooling issue. On the basis of this result, the current condition of micro sample and census data were utilized to compare the IPF (Iterative Proportional Fitting) of previous methodology with the establishment result of suggested algorithm. It is found that the algorithm is valid and built with the synthetic population based on SA through statistical verification. It is resulted from the application of the suggested SA method that such tradition algorithm as IPF has zero-cell or sample biased problems. However, not only can SA algorithm overcome such problems but also it can effectively address hill climbing and cooling schedule issues.