Inferring hybrid transportation modes from sparse GPS data using a moving window SVM classification

► Inferring transportation mode from GPS data without user information or context assumptions. ► Speed and acceleration are the variables with the highest mode discriminatory power. ► Transportation mode inference is based on a moving window SVM classification. ► GPS trips segmented into single-mode stages to preserve classification cohesiveness. ► Our model achieves 88% mode classification accuracy with both variables combined.