Using geometric and semantic attributes for semi-automated tag identification in OpenStreetMap data

OpenStreetMap is one of the successful volunteered geographical information projects. Participants contribute to this crowdsourced project by adding geometric and semantic data. However, both missing geometric and semantic data still cause completeness problems. In this paper, a semi-automated approach is suggested to identify the values of leisure tag of polygon features. The approach uses geometric (rectangularity, density, area, and distances to bus stop and shop) and semantic (amenity) data and estimates the key values using random forest classifier. In short, the results show that tag identification was conducted in three districts of Ankara with f-scores 78%, 86%, and 87%.