Using Smartphones to Perform Transportation Mode Determination at the Trip Level

In this paper, the authors address the problem of determining the transportation mode at the trip level using mobile phones. The authors describe an algorithm that processes raw global positioning system (GPS) and accelerometer data to detect whether a person is standing, walking, biking, driving or taking public transit and further processes this data to determine the trip mode. This work contributes to existing research, which determines exact transportation modes at each time instant, by adding context to the mode data with trip definitions that are defined by the activity-based travel model specification for the San Francisco Bay Area. The authors describe the features extracted from GPS and accelerometer sensors used to classify modes and trips with machine learning algorithms. The authors present a test of the algorithm conducted in the San Francisco Bay Area consisting of 28 trips.