A Dynamic Choice Model with Heterogeneous Decision Rules: Application in Estimating the User Cost of Rail Crowding

摘要: Crowding valuation of subway riders is an important input to various supply-side decisions transit operators. The crowding cost perceived by a rider generally estimated capturing the trade-off that makes between and travel time while choosing route. However, existing studies rely on static compensatory choice models fail account for inertia learning behaviour riders. To address these challenges, we propose new dynamic latent class model (DLCM) which (i) assigns inertia/habit classes based different decision rules, (ii) enables transitions over time, (iii) adopts instance-based theory We use expectation-maximisation algorithm estimate DLCM, most probable sequence each retrieved using Viterbi algorithm. proposed DLCM can be applied in any context capture dynamics rules used decision-maker. demonstrate its practical advantages estimating Asian metro's calibrate model, recover daily route preferences in-vehicle experiences regular metro two-month-long smart card vehicle location data. results indicate average follows rule only 25.5% occasions. estimates also show increase 47% riders' under extremely crowded conditions relative uncrowded conditions.