The ability to predict the choices of prospective passengers allows airlines to alleviate the need for overbooking flights and subsequently bumping passengers, potentially leading to improved customer satisfaction. Past studies have typically focused on identifying the important factors that influence choice behaviors and applied discrete choice framework models to model passengers’ airline choices. Typical discrete choice models rely on two major assumptions: the existence of a utility function that represents the preferences over a choice set and the linearity of the utility function with respect to attributes of alternatives and decision makers. These assumptions allow the discrete choice models to be easily interpreted, as each unit change of an input attribute can be directly translated into change in utility that eventually affects the optimal choice. However, these restrictive assumptions might impede the ability of typical discrete choice models to deliver operational accurate prediction and forecasts. In this paper, we focus on developing operational models that are intended for supporting the actual prediction decisions of airlines. We propose two alternative approaches, pairwise preference learning using classification techniques and ranking function learning using evolutionary computation. We have empirically compared these approaches against the standard discrete choice framework models and report some promising results in this paper.