A variational autoencoder approach for choice set generation and implicit perception of alternatives in choice modeling

Choice set generation is a challenging task, since the consideration generally unknown to modelers, and full choice could be too large enumerated. The proposed variational autoencoder approach (VAE) motivated by idea that chosen alternatives must belong set. VAE method explicitly considers maximizing likelihood of including in inferring underlying process. This paper derives generalized extreme value (GEV) model with implicit availability/perception (IAP) alternatives, for bridging modeling. Specifically, cross-nested logit (CNL) IAP derived as an example IAP-GEV models. assumes each alternative associated degree (likelihood context VAE) included route illustrated toy network. Simulation experiments show reproduce pre-defined true values. We further exemplify using real dataset. IAP-CNL estimated has best performance terms goodness-of-fit prediction performance, compared multinomial models conventional methods.