Unveiling Trips—A LABDA Consortium Study

Document Type

Article

Publication Date

12-2025

Identifier

DOU: 10.1123/jmpb.2025-0009

Abstract

Background: Understanding human movement behavior is essential for transportation planning, public health, and health geography research. To conduct movement behavior research, researchers need efficient methods to process data sets reliably and ensure validity. Therefore, we present a validation analysis of an innovative algorithm for detecting participants’ trips and transportation modes, such as walking/running, bicycling, or vehicle. Material and Methods: A modular algorithm to obtain the participants’ trips and transportation modes was used and compared with an annotated data set of 40 adults who wore a Global Positioning System logger (Qstarz), accelerometer (ActiGraph), and wearable camera (SenseCam) for a mean of 4 days. Pictures from the SenseCam were annotated to classify whether each 2-min period was part of a trip and, if so, which transportation mode was used (walking/running, bicycling, or vehicle). Commonly used classification metrics were calculated for 15-s epochs. Results: The trip classification precision was 79%, with a sensitivity of 84% and an F1-score of 81%. The accuracy was 93%. Transportation mode classification performed best for bicycling (>90% for all metrics), followed by vehicles (>85%), and walking/running (>75%), with an accuracy of 90%. Conclusion: Our modular algorithm has good validity for detecting trips and identifying transportation modes in population studies. Its implementation is readily available as a Python package, making it easy for researchers to use.

Journal Title

Journal for the Measurement of Physical Behaviour

Volume

8

Issue

1

Keywords

accelerometers; classification; GPS; GNSS; physical activity; transportation

Link to Full Text
Library Record

Share

COinS