Abstract
Ridesharing recommendation is an important application in urban computing. The existing grid map method is a popular method but may overlook many possible ridesharing opportunities. In this paper, we proposed an algorithm to find ridesharing paths that consist of two stages. In the first stage, GPS tracjectoreis are segmented and represented as cubes, and in the second stage, those cubes serve as landmarks for identifying possible ridesharing paths. We used the GeoLife GPS trajectories dataset to evaluate this approach and compared our algorithm with the grid map method. The results show that the number of possible ridesharing paths identified by our approach is six times that of the grid map method.