From Tweets to Semantic Trajectories: Mining Anomalous Urban Mobility Patterns

Abstract

This paper proposes and experiments new techniques to detect urban mobility patterns and anomalies by analyzing trajectories mined from publicly available geo-positioned social media traces left by the citizens (namely Twitter). By collecting a large set of geo-located tweets characterizing a specific urban area over time, we semantically enrich the available tweets with information about its author – i.e. a resident or a tourist – and the purpose of the movement – i.e. the activity performed in each place.

We exploit mobility data mining techniques together with social network analysis methods to aggregate similar trajectories thus pointing out hot spots of activities and flows of people together with their variations over time. We apply and validate the proposed trajectory mining approaches to a large set of trajectories built from the geo-positioned tweets gathered in Barcelona during the Mobile World Congress 2012 (MWC2012), one of the greatest events that affected the city in 2012.