University of California Transportation Center

The goal of this dissertation is to develop an activity-based trip generation model which addresses shortcomings of the conventional trip-based approach. Problems with conventional generation models resulted from a fundamental incapability to address the temporal and spatial characteristics of activities and the trips which they generated. The sequencing and scheduling of trips and activities, and interactions between household members, are ignored in the standard model. The proposed activity-based generation model was developed to estimate trip production from the analysis of complete travel/activity patterns. This approach classifies travel patterns with respect to activity, spatial, and temporal characteristics; standard trip rates can be also estimated from these representative activity patterns. In addition to a standard category production model, a stochastic logit-based pattern choice model and a deterministic discriminant analysis model were developed to stimulate activity pattern choice and the associated trip production level. A variety of variables describing the socioeconomic and demographic attributes at the household or personal level comprise the utility functions for choice prediction. Temporal stability of activity patterns was evident in similar life cycle groups in the 1985 and 1994 Portland test data, supporting the conclusion that patterns are a viable structure on which to base future forecasts.