Prior to the FIFA World Cup Qatar 2022 \(^\textrm{TM}\) , we conducted a survey regarding the use of different modes of transportation. This empirical research was performed to calibrate a passenger demand model to estimate the number of passengers at metro stations, considering a no-show rate and providing a basis for admission control. Decision theory and machine learning techniques were used to identify the key factors that influence transport choices. Interaction terms and binary variables from a decision tree created by the CART algorithm were included to capture non-linear effects. Maximum likelihood with a lasso penalty term was applied to estimate the high-dimensional utility function of the multinomial logit model, resulting in a sparse utility function. Our results illustrate that the use of a data-driven method can provide remarkably accurate predictions for transport mode choice analysis. This can be achieved without time-consuming and subjective preprocessing. Most importantly, no additional expert knowledge is required.