LDA model for privacy access based on trajectories and POIs

A privacy access method is proposed by combining trajectories and POIs via LDA model. Firstly, the Mean-Shift clustering algorithm is used to integrate individual user travel trajectories, combining with POI information to access the user’s latent travel patterns and other privacy. Then, a latent Dirichlet allocation (LDA) analysis is conducted on the travel patterns of all users in the dataset, constructing an LDA model that classifies users based on the features exhibited by their travel patterns. The model is used in practical applications to extract some of the sensitive attributes of users without resorting to other user information, and the ablation experiments show that the model has high accurate and precise access.