Detecting Activity Types and Trip Purposes from Passive GPS Data: A Data Mining Approach

Applying GPS (Global Positioning System) and other positioning technologies in a passive way has become a promising method for collecting individuals’ activity-travel behavior data due to its minimum burden on respondents. A major obstacle of such applications is to derive activity-travel behavior information such as activity types and trip purposes from the GPS data. In the past years, much research effort has been spent to detect activity types and trip purposes from passive GPS data. However, no commonly used method can be identified in the literature. This study proposes a data mining approach. Specifically, we develop a genetic algorithm to detect activity types and trip purposes through mining GPS tracking data, land use data, and socio-economic information. This algorithm has good self-learning and self-adaptation capabilities and needs neither prior knowledge nor artificial interference in the process of searching for the optimal solutions. The field study was conducted in Guangzhou, China and data were collected to test the applicability of the algorithm and the data mining approach. The results show that though constrained by data availability, the algorithm and data mining approach proposed in this study can detect activity types and trip purposes with accuracy rates reasonably good and comparable to that of other studies.