Sparse Trajectory Prediction Methods Based On Tensor Factorization

With the rapid development of mobile communication technologies,Global Positioning System technology and smart phones,the trajectory data of mobile objects has an explosive growth,the demand for trajectory prediction of moving objects is gradually increasing,and the trajectory prediction of moving objects has become the current research.This thesis takes sparse trajectory prediction as a research issue,and focuses on sparse trajectory prediction.The main contents are as follows.(1)Entropy-based Sparse Trajectories Prediction Enhanced by Matrix FactorizationExisting moving object's trajectory prediction algorithms suffer from the data sparsity problem,which affects the accuracy of the trajectory prediction.Aiming to the problem,we present an entropy-based sparse trajectories prediction method enhanced by matrix factorization(ESTP-MF).Firstly,we do trajectory synthesis based on trajectory entropy and put synthesized trajectories into the trajectory space.It can resolve the sparse problem of trajectory data and make the new trajectory space more reliable.Secondly,under the new trajectory space,we introduce matrix factorization into Markov models to improve the sparse trajectory prediction.It uses matrix factorization to infer transition probabilities of the missing regions in terms of corresponding existing elements in the transition probability matrix.It aims to further solve the problem of data sparsity.Experiments with T-Drive taxi trajectories dataset show that ESTP-MF generally improves prediction accuracy compared to the Baseline algorithm,Sub Syn algorithm and STP-EE algorithm.(2)Fast Time-aware Sparse Trajectories Prediction with Tensor FactorizationAs the markov model of trajectory prediction use a single transition probability matrix,which neglects the “time awareness" lead to low accuracy.Aiming to the problem,we propose a fast time-aware sparse trajectories prediction with tensor factorization method(TSTP-TF).We introduce multidimensional tensor modeling into Markov model to add the time dimension.Tensor factorization is adopted to infer the missing regions transition probabilities to further solve the problem of data sparsity.Due to the scale of the tensor,we design a divide and conquer tensor factorization model to reduce memory consumption and speed up decomposition.Experiments with real dataset show that TSTP-TF improves prediction accuracy compared to the ESTP-MF.(3)Sparse Trajectories Prediction based on Differential Privacy ProtectionThe trajectory data may leak personal privacy information.The Markov probability transition matrix is used as the intermediate data of the trajectory prediction.Differential privacy protection can reduce privacy disclosure.We focus on how to reduce the risk of privacy disclosure when publishing and using the markov probability transition matrix.Then we discuss the predictive usability of sparse trajectory prediction methods under differential privacy protection.Firstly,based on the Laplacian noise mechanism,we propose a differential privacy protection method based on transition state number restrictions(DPP-TSNR).Then,in the case of releasing data under differential privacy protection,we use the sparse trajectory prediction method based on matrix factorization to make trajectory prediction,so as to improve prediction accuracy and complete sparse trajectory prediction under differential privacy protection.Experiments with the Gowalla dataset and the Geo Life dataset show that DPP-TSNR can perform differential privacy protection well.DPP-TSNR ensures the privacy of trajectory sparse data while improving data availability.Under differential privacy protection,our method greatly improves the prediction accuracy compared with the trajectory prediction method based on markov model.