Southern California Association of Governments (SCAG) started to develop an Activity-Based Model (ABM) in 2009. To ensure that SCAG’s ABM reasonably replicates observed travel patterns, rigorous model calibration was performed after the model estimation step. Although calibration is an essential step in examining how well model components match some measurement of travel patterns, it cannot evaluate whether model estimation results are reasonable and thus adjustments to model parameters may not correct underlying issues with model estimation.
ABM is a new generation of travel demand model. It has many sub-models that are typically estimated with advanced techniques and involves many input variables. It is a challenge for modelers to fully comprehend how each estimated model is reasonably linked to travel behavior. In addition, given constrained budget and schedule, ABM consultants may not have sufficient resource to conduct detailed analysis for each sub-model. To resolve this issue, led by in-house modelers, SCAG adds a model assessment step after model estimation.
The new model assessment procedure will be tested on the work activity generation and workers’ travel scheduling models, which are important model components of SCAG’s ABM. Sub-models for work activity generation include work activity decision, work start/end time, and work duration. Workers’ travel scheduling models include tour mode choice, stop frequency and activity sequence, stop duration, travel time to stop, stop location, and stop mode choice.
The first objective of this study is to introduce the framework and procedure for SCAG’s ABM Model Assessment Step. The second objective of this study is to show the comparative model assessment results for all of the sub-models described above. The relationship between each model target (dependent variable) and the following factors will be analyzed: individual socioeconomic attributes, working status and arrangement, and land use/built environment characteristics and accessibility for both residential and work location.
The expected contributions of this study are: 1) the model assessment as a useful procedure for ABM development/enhancement projects, and 2) refined work activity analysis to help modelers better understand workers’ activity decision and travel behavior.