Research On User Movement Prediction Algorithm For LBS

Location-based services(LBS)generate a huge amount of trajectory data,which provides valuable data support for understanding user movement.Based on these data,this thesis studies an efficient algorithm for predicting the next location given the current trajectory segment and its associated information.Although a variety of solutions have been proposed in academia,several key challenges remain to be addressed in order to better predict human mobility:(1)Sociality: How to incorporate complex social relationships between users in trajectory prediction algorithms;(2)Periodicity: How to effectively improve the performance of the prediction algorithm by multi-level periodicity that exists.This thesis mainly studies the user movement prediction algorithm for LBS based on the above problems.On the one hand,the social network and mobile trajectory data information are analyzed and mined to train an appropriate prediction model to achieve the next location prediction.On the other hand,considering the limitations of social relationships and using periodicity as a supplement,through the introduction of attention mechanisms to capture periodic information of the mobile trajectory to improve performance for next location prediction.The main work and research completed are as follows:(1)The next location prediction algorithm integrating social networks is proposed.First,the relationship between social relationships and user movement is verified by data analysis on real data sets.Based on this,Word2 vec is used to model the social network to obtain the user vector.Then the LSTM is used to model the trajectory to obtain the trajectory vector.Finally,the social network and the trajectory are combined into a final model of the chapter,thereby improving the accuracy and operational efficiency of the model.According to the experimental results,the model of this model is better than other methods,and the addition of social relationships can improve the accuracy of prediction.(2)The next location prediction algorithm based on attention is proposed.Firstly,through the data analysis in the real data set,On the one hand,it discovers the limitations of social relationships on mobility and the necessity of periodicity as a supplement,and on the other hand verifies the relationship between periodicity and user movement.This chapter uses the LSTM and attention modules to capture sequential and periodic information in the user's current and historical trajectories.Combined with the social network to form our final model,the experimental results show that the next location prediction algorithm based on attention can further improve the prediction performance,indicating that the introduction of the attention mechanisms can improve the accuracy of the prediction.In general,the model designed in this thesis can make full use of the sociality and periodicity of user movement to improve the performance of prediction.A large number of experiments on real data sets show that the proposed algorithm has better performance.