Richard Wen, Claus Rinner
rwen@ryerson.ca, crinner@ryerson.ca
A short peer-reviewed paper and poster for the Ninth International Conference on Geographic Information Science in Montreal, Canada from September 27, 2016 to September 30, 2016.
OpenStreetMap (OSM) data consists of digitized geographic objects with semantic tags assigned by volunteer contributors. These human and machine readable tags are edited manually and automatically to improve data quality. The structure of the tags allow machine learning algorithms to support user editing by learning to identify irregular objects and data patterns. This research experimented with a random forest algorithm on geospatial variables for geospatial outlier detection and knowledge discovery in OSM data without ground-truth reference data.
We would like to thank the Geothink Social Sciences and Humanities Research Council (SSHRC) Partnership Grant for the funding provided during the duration of this research. Map data copyrighted OpenStreetMap contributors and available from http://www.openstreetmap.org
The code used in this short paper was developed for a Masters thesis available at github.com/rrwen/msa-thesis