Nivan Ferreira, James Klosowski, Carlos Scheidegger, Claudio Silva. Eurovis 2013, Computer Graphics Forum, v. 32, p. 201-210, 2013. Honorable mention for best paper award.
Scientists study trajectory data to understand trends in movement patterns, such as human mobility for trafﬁc analysis and urban planning. In this paper, we introduce a novel trajectory clustering technique whose central idea is to use vector ﬁelds to induce a notion of similarity between trajectories, letting the vector ﬁelds themselves deﬁne and represent each cluster. We present an efﬁcient algorithm to ﬁnd a locally optimal clustering of trajectories into vector ﬁelds, and demonstrate how vector-ﬁeld k-means can ﬁnd patterns missed by previous methods. We present experimental evidence of its effectiveness and efﬁciency using several datasets, including historical hurricane data, GPS tracks of people and vehicles, and anonymous cellular radio handoffs from a large service provider.