The distribution of transit stations constitutes an ubiquitous task in large urban areas. In particular, bus stops spacing is a crucial factor that directly affects transit ridership travel time. Hence, planners often rely on traffic surveys and virtual simulations of urban journeys to design sustainable public transport routes. However, the combinator-ial structure of the search space in addition to the time-consuming and black-box traffic simulations require computationally expensive efforts. This imposes serious constraints on the number of potential configurations to be explored. Recently, powerful techniques from discrete optimization and machine learning showed convincing to overcome these limitations. In this preliminary work, we build combinatorial surrogate models to approximate the costly traffic simulations. These so-trained surrog-ates are embedded in an optimization framework. More specifically, this article is the first to make use of a fresh surrogate-assisted optimization algorithm based on the mathematical foundations of discrete Walsh functions in order to solve the real-world bus stops spacing optimization problem. We conduct our experiments with the sialac benchmark in the city of Calais, France. We compare state-of-the-art approaches and we highlight the accuracy and the optimization efficiency of the proposed methods.