Land use discovery based on Volunteer Geographic Information classification

Abstract

Nowadays, cities are dynamic ecosystems where urban changes occur at a very fast pace. Hence, social sensing has become a powerful tool to uncover the actual land-use of a metropolis. However, current solutions for land-use discovery based on user-generated data usually rely on an information retrieval mechanism applied on a textual corpus. This causes ad-hoc place labelling with limited semantic meaning. In this line, the present work introduces a novel data-driven methodology that extends existing solutions by means of a classifier based on a pre-defined hierarchy of land categories. Two types of social networks –text-based and venue-based platforms– are utilized to train the classifier, which is then applied to infer the use of the land based on text data in areas where venue data are not available. The approach has been evaluated by using large datasets comprising two large cities, showing an accuracy above 90% in predicting the land-use categories.