Floating Car Data mining: Identifying vehicle types on the basis of daily usage patterns
Résumé
This paper presents a novel approach for exploring vehicle daily mobility patterns using Floating Car Data in the Paris region. The objective is to reveal vehicle types and analyze usage patterns on different roadway sub-classes. Firstly, mobility representative features are recovered to recognize vehicle activity contexts, which includes modeling single trip pattern in terms of time-window, traveling-distance, and speed; and building vehicle mobility profile of trip combinations. Based on that, a twostep clustering algorithm is developed to explore the constitution of trip patterns and cluster vehicle types. Characterization analysis is conducted to find out outstanding features of clustered groups, thus helping to categorize the vehicles behavioral types. As a result, 4 major vehicle types were identified over the Paris region with the 2 comparative leading groups as those mainly composed by morning-activity trips and long-distance trips. Then, statistical association assessment by Configural Frequency Analysis is employed to examine the usage intensity on different roadway classes of identified types. The association analysis reveals that the identified types have statistically significant differences in the usage of different roadway classes. Furthermore, this approach can be expected to provide more representative results with more generalized sampling by the development of connected vehicles.
Origine : Fichiers éditeurs autorisés sur une archive ouverte