Research On Online Taxi Travel Destination Prediction Method Based On Long Short-Term Memory Network

With the growing development of online taxi platforms,the travel way of the public is becoming more and more intelligent,and in order to further improve the quality of travel,the issue of online taxi travel destination prediction has received attention from various aspects,but the trajectory destination prediction poses a threat to the location privacy security of users.The protection the privacy of users and the effective prediction of online taxi travel destinations has become urgent problems.To address the above problems,this paper conducts an in-depth analysis and research on the prediction of online taxi travel destinations in terms of the effectiveness of trajectory privacy protection,the usability of trajectories after privacy protection and the accuracy of destination prediction,the main contents are as follows:(1)A privacy protection model of online taxi trajectory based on generating adversarial networks is proposed.Firstly,a novel encoding method is proposed to encode trajectories and other additional information into specific numerical representations as input of the model which better solves the problem of incomplete utilization of spatial-temporal information;Secondly,A trajectory privacy-preserving models based on generative adversarial networks(LSAN)are used to generate synthetic trajectories with privacy protection effects,and they are published and shared as real trajectories.The deep learning models effectively resist the problem of knowledge background of the attacker;Finally,several experiments were conducted using real trajectory data sets to verify the effectiveness of this model.Compared with other existing privacy protection methods,this model has better privacy protection effects,and also has advantages in data availability.(2)A mobile trajectory destination prediction model based on long and short-term memory networks is proposed.For the trajectory characterization,a distributed representation method of trajectories is proposed.Firstly,the trajectory is meshed,and the high-dimensional unique heat code vector representing the position is reduced to generate low-dimensional embedded vectors containing geographical topological relationship.Secondly,the destination is clustered,and the cluster center is used as the label of the trajectories to reduce the difference of similar trajectories and enlarge the characteristics of different trajectories,effectively overcoming the problem of data sparsity.In the destination prediction,the self-attention mechanism is introduced into the long short-term memory network.The key points in the sequence are mined and the weights are allocated according to their importance,which better solves the long-term dependence problem.Finally,many experiments were carried out on the real trajectory data sets to verify the effectiveness of the model.And compared with the existing models,the proposed model has better accuracy.(3)On the basis of the above research,a prototype cab destination prediction system for location privacy protection is designed and implemented using Python language,Python GUI and Py Charm.Detailed analysis is carried out from the aspects of demand analysis,structural system and software functions.The system operation results show that the online reservation destination prediction prototype system for location privacy protection can accurately predict the travel destination under the premise of effectively protecting the customer’s location privacy,providing effective technical support for customer security and location-based services.