Detecting Travel Modes and Profiling Commuter Habits Solely Based on GPS Data

The Global Positioning System (GPS) has been gaining importance for travel surveys since the 1990s. While it is successfully used to collect accurate information about traveled routes and travel times, only little is known on extracting added information like transport modes and trip purposes. In this paper a system for collecting and profiling commuter data is presented. This includes all steps from determining the right GPS device to processing the GPS trajectories using a data driven approach without relying on geographic information systems (GIS) or user input. A participant’s habit in route choice, travel times, travel mode changes and travel modes is extracted and stored in a profile. Using this, at the beginning of a commute, routes that the commuter is likely to take are determined and used to provide personalized real-time time traffic information. Extracting the commuter profile includes pre-processing methods for the trajectories, detection methods for places of travel mode changes and segmentation of tracks into singular travel modes. Also a detailed mode detection step is performed, comparing decision trees, logistic regression, multilayer perceptions and support vector machines as classification methods. Finally, the data is reduced to a network of prototypes using Growing Neural Gas (GNG), a self-organizing-network algorithm. This in turn enables a more effective algorithm for detecting likely routes. The mode detection algorithms achieved a detection rate of about 84% on a test sample, while the commuting-routes were predicted correctly 80% of the time within five minute from the start of the commute.