Decision Tree Approach to Predicting Vehicle Stopping from GPS Tracks in a National Park Scenic Corridor. Issue 3 (March 2019)

• Record Type: Journal Article
• Authors: Fuentes, Antonio; Heaslip, Kevin; Sisneros-Kidd, Abigail M.; D'Antonio, Ashley; Kelarestaghi, Kaveh Bakhsh
• Is Part Of: Transportation research record. Volume 2673:Issue 3(2019)
• Page Start: 86
• Page End: 96
• Publication Date: 2019-03
• Subjects: Transportation -- Periodicals; Roads; Transport -- Périodiques; Routes -- Périodiques; Routes -- Conception et construction -- Périodiques; Roads; Transportation; 388.05
• DOI: 10.1177/0361198119835515
• Languages: English
• ISSNs: 0361-1981

Abstract

In this study, a GPS tracking dataset was utilized to predict the probability of a vehicle stopping along a scenic corridor in a national park setting. The Moose-Wilson Corridor (MWC) in Grand Teton National Park was evaluated for vehicle stopping/visiting determined from GPS collected data along the corridor. Four attractions were evaluated, which consisted of Death Canyon, Granite Canyon, the Laurance S. Rockefeller Preserve, and Sawmill Ponds. A decision tree analysis was implemented to determine the probabilities of visitors' stopping patterns at park attractions. A benefit to the decision tree method is the easy to read structure and simple visual representation. An 814-observation sample set was split into training and testing datasets that resulted in model accuracies as high as 94%. A promising outcome of this methodology is that it provides a visual reference to help identify attraction relationships. Similarities in the tree structure were determined for Death Canyon and Granite Canyon owing to their child nodes being composed of other MWC attractions. Alternatively, the Laurance S. Rockefeller Preserve and Sawmill Ponds attractions determined more variability in their tree structure. The implementation of a data analysis method such as the one presented in this paper could help national park managers prepare for the incoming era of technology and data.