Statistical characteristics and transportation mode identification of individual trajectories

Abstract

Entering big data era, individual GPS trajectory data has created great opportunities for human mobility and collective behavior studies. Individual GPS trajectories can be collected by location-based services on mobile phones. However, GPS data often do not record transportation modes (e.g., walking, riding a bus, or driving a car). In this study, we analyzed the statistical characteristics of individual trajectories and present a collaborative isolation forest (Co-IF) model to identify the transportation modes of mobile phone GPS trajectories. Unlike previous models that identify multiple transportation modes simultaneously, the proposed Co-IF model builds a single-class classifier for each transportation mode and then combines their results. Compared to existing models, the Co-IF model offers competitive performance and shows improved reliability with noisy trajectories.

Publication
International Journal of Modern Physics B, vol. 0(0) pp. 0, February 2020.