Abstract:
Microsimulation approaches to travel demand forecasting are gaining increasing attention due to their ability to replicate the multitude of factors underlying individual travel behavior. The implementation of microsimulation approaches usually entails the generation of synthetic households and their associated activity-travel patterns to achieve forecasts with desired levels of accuracy. This paper develops a sequential approach to generate synthetic daily individual activity-travel patterns. The sequential approach decomposes the entire daily activity-travel pattern into various components, namely, activity type, activity duration, activity location, work location, and mode choice and transition. The sequential modeling approach offers practicality, provides a sound behavioral basis, and accurately represents individual's activity-travel patterns. In the proposed system, each component may be estimated as a multinomial legit model. Models are specified to reflect potential associations between individual activity-travel choices and such factors as time-of-day, socio-economic characteristics, and history dependence. As an example, the paper furnishes results for activity type choice models estimated and validated using the 1990 Southern California Association of Governments travel diary data set. Validation results show that the predicted pattern of activity choices conforms with observed choices by time of day. Thus the paper shows that realistic daily activity-travel patterns, which are requisites for microsimulation approaches, can be generated for synthetic households in a practical manner.
