The widespread use of mobile phones has generated a large amount of individual trajectory data. Such
data can greatly help analyze and understand human daily travel behavior. In this paper, we use the topic modeling
technique to infer trip purposes based on pseudonymized users’ trajectory data from cellular network and points of
interest (POIs) from online map services. The adapted latent Dirichlet allocation method is used to model the trip
generation process and then infer trip purposes behind the data. The experiments are performed on a data set of
27,732 trip records in Beijing on weekdays. Ten topics are discovered. This method can easily infer different trip
purposes based on three trip attributes, i.e., trip departure time, stay duration, and POI categories for destinations,
and most of the topics/trip purposes are explainable.