A-TPN: User trajectory prediction method based on mobile communication CDR data

Existing predictions of human behaviour trajectories in mobile communication scenarios are mainly based on rules and Markov statistical methods. At the same time, less research has been conducted on CDR data using deep learning methods. This paper proposes the A-TPN (Adaptive Trajectory Prediction Network) algorithm to predict human trajectories for the large scale and sparsity of Call Detail Records (CDR) data. Experiments show that A-TPN achieves a prediction accuracy of 94.76% on real datasets, which is 8.7% better than the traditional Markov statistical computation method. This paper demonstrates the algorithm's effectiveness through rich comparative experiments and analysis.