Understanding the tourist travel patterns is essential for tourist management and planning. This study proposes a novel three-step framework to extract popular tourist routes from large-scale GPS data. In the first step, tourist trajectories are identified from original GPS data based on the distance between GPS trajectory points and the point of interest. We gridded the study area to facilitate this matching process. In the second step, dynamic time wrapping is applied to measure the distance between tourist trajectories from the spatial dimension, and earth mover’s distance is applied to measure the distance from the temporal dimension. We propose a novel similarity index to combine these two distances and measure the spatial-temporal similarity between tourist trajectories. In the third step, the popular tourist routes are obtained by clustering tourist trajectories with high similarity index. We validated the proposed framework with a real-world GPS dataset collected in Beijing, China, and conducted the sensitivity analysis of important parameters. The extracted popular tourist routes are consistent with the actual popular tourist routes, suggesting that our framework effectively extract meaningful tourist route information from large-scale GPS data.

GPS data, Popular Tourist Routes, Tourist Identification, Spatial-Temporal Similarity Analysis