Semantically enriching personal mobility data using OpenStreetMap: a case study using smartphone users' frequently...

Area of Science:

• Geographic Information Science
• Behavioral Science
• Data Science

Background:

• Behavioral scientists require accurate point-of-interest (POI) labels for mobile device location data.
• Volunteered geographic information (VGI), like OpenStreetMap (OSM), offers cost-effective annotation but data quality is debated.
• Limited evidence exists on VGI's suitability for personal mobility data annotation.

Purpose of the Study:

• To assess the performance of OpenStreetMap (OSM) point-of-interest (POI) data for annotating personal mobility data.
• To compare OSM's annotation accuracy against participant-provided labels and commercial geodatabases (Foursquare, Google Maps).
• To provide empirical evidence for using OSM in the semantic enrichment of mobile location data research.

Main Methods:

• Developed a Python package to extract POI geometric and semantic data from OpenStreetMap (OSM).
• Annotated year-long smartphone location data from 93 participants in the United States using the OSM package.
• Evaluated annotation performance against participant-provided labels and compared with Foursquare and Google Maps.

Main Results:

• OpenStreetMap (OSM) data achieved the best overall performance in semantic annotation across eight place categories.
• OSM labeled 81% of places with an average F1 score of 0.65, demonstrating its utility for mobility data.
• While OSM excelled overall, Foursquare and Google Maps showed advantages in annotating specific place categories.

Conclusions:

• This study provides empirical support for utilizing OpenStreetMap (OSM) data for semantic enrichment in mobile location data research.
• OSM offers a cost-effective and efficient alternative to commercial databases for annotating personal mobility patterns.
• Recommendations are provided for future implementations of OSM in behavioral science research utilizing location data.