Understanding travel-activity behaviors is important for transport planning, particularly for the provisions of new mobility services. Exploratory analyses can be helpful in getting initial insights into these behaviors. This study presents an application of Graph Edit Distance (GED), which has yet to be explored in a transportation context, for clustering weekly Travel-Activity Patterns (TAPs). We used a dataset of 20,480 respondents in an online survey in Japan in 2021 for evaluating the approach. TAPs were represented as graph objects and a GED matrix was computed with the optimal edit cost structure obtained using a robustness index (RI). The K-medoids algorithm identified six clusters, including two clusters of car drivers, two clusters of public transport, one cluster of active travel modes, and one cluster of multiple travel modes. These clusters were visualized by a two-dimensional space produced by Discriminant Correspondence Analysis (DICA), with car access, occupation status, and age identified as the most important variables for discriminating them. Overall, the usefulness of GED (and the follow up DICA) in clustering TAPs was illustrated, and some directions for further uses of GED in studying travel-activity behaviors were discussed.
Travel-Activity Patterns Identification Using Graph Edit Distance
24.09.2023
893541 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
IEEE | 1999
|British Library Conference Proceedings | 1999
|Edit distance from graph spectra
IEEE | 2003
|Edit Distance from Graph Spectra
British Library Conference Proceedings | 2003
|Approximate graph edit distance computation by means of bipartite graph matching
British Library Online Contents | 2009
|