NTC/ITSMD Host Seminar Featuring the work of Dr. Sharif Azedeh

On Thursday November 5, Dr. Sharif Azedeh, from TRANSP-OR at EPFL, gave a speech to students and faculty from multiple backgrounds at the University of Maryland.

In the presentation, she addressed the problem of demand modeling from a new point of view when customers come from different segments with different willingness to pay and characteristics.

Dr. Azedeh proposed an algorithm that takes availability constraints into account via a non-parametric mathematical representation. To solve the problem, she introduced a modified CDLP model, with customer’s choice probabilities are derived from multi-nominal logit model. A mixed integer nonlinear formulation based on the estimated customer’s preference vector were also introduced to minimize the prediction error.

Unlike traditional choice models, product utilities are defined as variables in the model that are estimated only on historical data. As illustrated in the presentation, the method can be helpful to revenue management systems in transportation companies or any other service companies.

Published November 12, 2015