ONLINE DISCRETE CHOICE
Team: Mazen Danaf, Yifei Xie, Moshe Ben-Akiva
A framework is developed for estimating and updating user preferences in an app-based personalization system (e.g. recommendation, personalized pricing)
A Hierarchical Bayes estimator is proposed in order to account for inter- and intra-personal heterogeneity.
An online procedure is used to update user preferences in real time upon making a choice.
Discrete choice models have been widely applied in different fields to better understand behavior and forecast market shares. Because of their ability to capture taste heterogeneity, logit mixture models have gained increasing interest among researchers and practitioners. However, since the estimation of these models is computationally expensive, their applications have been limited to offline contexts. On the other hand, online applications (such as recommender systems) require users' preferences to be updated frequently and dynamically.
We develop a methodology for estimating discrete choice models online, while accounting for inter- and intra-consumer heterogeneity. An offline-online framework is proposed to update individual-specific parameters after each choice using Bayesian estimation. The online estimator is computationally efficient, as it uses the data of the individual making the choice only in updating his/her individual preferences. Periodically, data from multiple individuals are pooled, and population parameters are updated offline.
Online estimation allows for new and innovative applications of discrete choice models such as personalized recommendations, dynamic personalized pricing, and real-time individual forecasting. This methodology subsumes the utility-based advantages of discrete choice models and the personalization capabilities of common recommendation techniques by making use of all the available data including user-specific, item specific, and contextual variables.
In order to enhance online learning, two extensions are proposed to the logit mixture model with inter- and intra-consumer heterogeneity. In the first extension, socio-demographic variables and contextual variables are used to model systematic inter- and intra-consumer taste heterogeneity respectively. In the second extension, a latent class model is used to allow for more flexibility in modeling the inter- and intra-consumer mixing distributions.
The proposed methodology has been applied to Tripod and personalized tolling where it was shown that prediction performance improves over time.
Danaf, M., Becker, F., Song, X., Atasoy, B. and Ben-Akiva, M., 2019. Online discrete choice models: Applications in personalized recommendations. Decision Support Systems, 119, pp.35-45.
Danaf, M.M.S., 2019. Online discrete choice models: applications in smart mobility (Doctoral dissertation, Massachusetts Institute of Technology).
Danaf, M., Atasoy, B. and Ben-Akiva, M., 2020. Logit mixture with inter and intra-consumer heterogeneity and flexible mixing distributions. Journal of choice modelling, 35, p.100188.
Xie, Y., Zhang, Y., Akkinepally, A.P. and Ben-Akiva, M., 2020. Personalized choice model for managed lane travel behavior. Transportation research record, 2674(7), pp.442-455.
Danaf, M., Guevara, A., and Ben-Akiva, M. 2021. A Control-Function Correction for Endogeneity in Random Coefficients Models: The Case of Choice-Based Recommender Systems. Working paper.