An online updating method for time-varying preference learning

摘要: Abstract The rapid proliferation of smart, personal technologies has given birth to smart Transportation Demand Management (TDM) systems that can give personalized incentives users. This personalization capacity builds on accurate modeling user behaviors; however, in practice, a user’s behavior data is often limited, and his preferences the discrete choice-making process may change or evolve. In this paper, we propose new online-updating model accurately efficiently estimate an individual’s from choices. Our built concept canonical structure, where set models are identified as common preference patterns shared by whole population, membership vector also for each individual capture degrees resemblance those patterns. To allow vary process, time-varying be integrated with structure. current study, use simple cubic polynomial single variant show detailed formulation model. An strategy proposed, such it possible update parameters partially practice. proposed suitable heterogeneous population insufficient individual. Both simulation studies real-world application taken study. results comparing other frequently used models, highest accuracy learning prediction.