Abstract : This paper describes probability models for human travel, using latent factors learned from data. The latent factors represent interpretable properties: travel distance cost, desirability of destinations, and affinity between locations. Individuals are clustered into distinct styles of travel. The latent factors combine in a multiplicative manner, and are learned using Maximum Likelihood. The resulting models exhibit significant improvements in predictive power over previous methods, while also using far fewer parameters than histogram-based methods. The method is demonstrated on travel data from two sources: geotags from a social image sharing site (Flickr), and GPS tracks from Shanghai taxis.
