Learning Latent Factor Models of Travel Data for Travel Prediction and Analysis

Abstract

Abstract. We describe latent factor probability models of human travel, which we learn 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. We show that our models explain the data significantly better than histogrambased methods. We also visualize the model parameters to show information about travelers and travel patterns. We show that different individuals exhibit different propensity to travel large distances. We extract the desirability of destinations on the map, which is distinct from their popularity. We show that pairs of locations have different affinities with each other, and that these affinities are partly explained by travelers' preference for staying within national borders and within the borders of linguistic areas. The method is demonstrated on two sources of travel data: geotags from Flickr images, and GPS tracks from Shanghai taxis.

A method for statistical analysis of repeated residential movements to link human mobility and HIV acquisition

“…Other important classes of statistical models for human mobility are Lèvy flights models [4] and multiplicative latent factor models [5]. Lèvy flights models make use of a power law to represent the probability that an individual changes their residence over a certain distance.…”

Loglinear model selection and human mobility

“…This represents a serious limitation since it implies that traveling to destinations that are located at the same distance from an origin is equally likely. A more recent contribution [38] builds on multiplicative factor models from social network analysis [41] to improve the Lèvy flights model which lacks the ability to quantify the desirability of certain travel locations. They propose a model in which P ij depends of a function f pd ij , τ q of distance d ij and of location-specific latent factors u i P R q and v j P R q : P ij 9 exp `f pd ij , τ q `uT i v j ˘, where u T i v j represent the affinity of locations i and j.…”

“…Both the Lèvy flights models [10] and the multiplicative latent factor models [38] are based on the crude assumption that human travel can be seen as a Markov process in which the probability of traveling to a location depends only on the origin of the trip's segment, and does not depend on previous locations visited. However, individuals are likely to travel repeatedly across multiple locations in a given period of time.…”