Answers are grounded in this paper and cite their sources.
The knowledge of mobility in a territory is essential for local authorities’ decision making. The multiplication of sensors in smart-phones is an important and reliable data source to analyze users’ travels. This paper presents a method for collecting GNSS and accelerometer data and a processing workflow to classify users’ transportation mode with high accuracy. Recommendations for collecting the measurements are explained. The pre-processing method presented is based on the analysis of time gaps between successive observations. The fusion of GNSS and accelerometer data and the calculation of features are performed with a sliding time window. OCC-Transport Mode, a dataset collected with a smartphone application is presented and made public to illustrate the different steps. The performances of several classifiers are compared on the collected dataset and on two public datasets (GeoLife and US-Transportation Mode). The classification accuracy, improved by the joint use of GNSS and accelerometer, is close to 100% on the collected dataset. The features resulting from the time gaps are more important than the other features in the classification. The results obtained on two public datasets are discussed.
No takes yet. Share an insight, caveat, or question.
Fourez et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: