THE STAGA-DATASET: STOP AND TRIP ANNOTATED GPS AND ACCELEROMETER DATA OF EVERYDAY LIFE
- 1Quality and Usability Lab, Berlin Institute of Technology, Berlin, Germany
- 2Institute of Medical Sociology and Rehabilitation Science, Charité - Universitätsmedizin Berlin, Berlin, Germany
- 3University of Applied Sciences Hamm-Lippstadt, Germany
- 4German Research Center for Artificial Intelligence (DFKI), Berlin, Germany
Keywords: GNSS, Dataset, Annotation, Accelerometer, Benchmark, Stop Trip Classification
Abstract. Identifying stops and trips from raw GPS traces is a fundamental preprocessing step for most mobility research applications. Thus, ensuring the excellent accuracy of such systems is of high interest to researchers designing such analysis pipelines. While there are plenty of GPS datasets available, these usually do not provide annotations and thus cannot be used for benchmarking stop/trip classifiers easily. This manuscript introduces a GPS & accelerometer dataset, including accurate stop/trip annotations. It contains 122,808 GPS samples as one continuous trajectory, spanning over 126 days. The recorded time frame includes working days, vacations, travelling, everyday life and all regular modes of transportation. During recording, a detailed mobility diary was conducted to capture each dwelling period’s exact beginning and end. The position and diary data combined contain 78,900 labelled stops and 43,908 labelled trips. This serves as ground truth for stop/trip classification algorithms to test existing tools or develop new analysis methods. The introduced dataset is freely available under a CC-By Attribution 4.0 International license, the annotation tool under the BSD 3-Clause license.