A Consistent Differential Privacy Dynamic Trajectory Flow Prediction Method

Ensuring privacy while maintaining accuracy in trajectory prediction is a crucial challenge in privacy-sensitive applications such as smart transportation and mobility analytics. This paper presents CDP-DTP (Consistent Differential Privacy Dynamic Trajectory Flow Prediction), a novel approach that effectively balances privacy protection and prediction accuracy in trajectory forecasting. The proposed method constructs a trajectory flow graph and integrates Laplace noise-based differential privacy with consistency constraint adjustments to enhance privacy while maintaining data utility. A CNN-LSTM hybrid model also extracts spatial and temporal features, improving prediction performance through feature fusion. Experiments on real-world trajectory datasets demonstrate that CDP-DTP outperforms traditional differential privacy methods by achieving lower mean squared error (MSE) while ensuring stronger privacy protection across different privacy budget settings. These results validate the model's effectiveness in privacy-sensitive trajectory prediction tasks. The proposed method provides a scalable solution for privacy-preserving mobility analytics and contributes to future research in intelligent transportation and secure data sharing.