Choice Set Generation for Modeling Scenic Route Choice Behavior with Geographic Information Systems

Abstract

Finding a scenic route between two locations is a common trip-planning task, in particular for tourists and recreational travelers. For the automated computation of a scenic route in a trip-planning system to function, it is necessary to understand which attributes of a route and its surroundings are associated with attractive scenery. One way to find this information is with a route choice model framework, which requires a choice set of realistic paths for traveling a scenic route. This paper proposes a constrained enumeration type algorithm for computing scenic alternatives with a branch-and-bound approach and a labeling algorithm in combination with link elimination to generate a scenic route choice set. These algorithms use volunteered geographic information—shared, geo-tagged photos (photos to which geographical identification metadata have been added) from Panoramio—to generate path alternatives. The results show that the proposed branch-and-bound algorithm replicated observe travel behavior better than the traditional deterministic minimum cost–based algorithms in terms of overlap and consistency. In addition, for the proposed branch-and-bound method, the average number of scenic clusters was almost identical to that found along observed scenic routes; this method outperformed all other evaluated algorithms for this aspect.