A Transit Network Analysis Under Random Regret Minimization

Abstract

The main objective of this research is developing path based stochastic traffic analysis frame on multi-modal urban transit network. Interest in the development and implementation of path-based stochastic traffic assignment (STA) models for traffic analysis is growing to explain complex decision making process. But path-based models are still emerging issue due to it was very hard to solve mathematical models. To challenge this objective, this research set several milestones. This research described semi-compensatory heuristic analysis frame on urban transit with the concept of regret in behavioral economics. It consist of path set generation stage and assignment stage. The parameters and result will be calibrated and verified with real world observation. Smart card data will be a good data source of real world behavior. Real scale network of multi-modal urban transit is tested to present real size network practicability with Seoul metropolitan area network. This research presented several points into path based transit analysis frameworks. First, the research introduced relative behavioral constraint thresholds rely on regret theory to explicit path set problem to averse extreme cases and to provide flexibility of model than absolute pre-fix value. the result presented how to extreme cases among similar path sets can be avoided rely on regret theory and threshold function. Second, regret based assignment (Random Regret Minimization) on transit network with generated path set was performed to present avoiding unconsidered paths improves assignment result. Assignment results are improved rely on explicit path set generation. Third, the research calibrated and verified the algorithm with sufficient observations from big amount of smart card data. Path set generations threshold function was estimated from sufficient observation. Due to regret theorys alternation relative flexibility and estimated threshold function there are no need to calibrate thresholds case by case.