Research On User Moving Trajectory Prediction

In recent years,urbanization has been further accelerated.The number of people and vehicles in cities has risen sharply,resulting in the growing demand for precise Location Based Services(LBS).In order to provide users with more tailored services,it has become a hot research area of the spatial-temporal database to mine and analyze users' moving preferences in recent years.At present,a large number of traffic trajectory data can be collected by enterprises and stored persistently.Trajectory data not only records the real-time location of users,but also contains rich potential characteristics of users' movement.This provides unprecedented opportunities for the prediction of users'moving trajectory.Therefore,based on massive traffic trajectory data,this paper deeply studies a more accurate technology of users' moving trajectory prediction.To predict users' moving trajectory,it is not only necessary to consider the data sparsity problem existing in the trajectory itself,but also the spatial-temporal evolution mode of trajectory.At the same time,traffic trajectory data usually contains contextual information.Users' behavior is vulnerable to external environment,such as weather,holidays and other factors.In addition,effective moving characteristics of users may appear in different spatial granularity,and traditional time-series modeling methods cannot effectively integrate these characteristics.To solve these complex problems,this paper constructs a new neural network model based on deep learning methods.The model has a high ability to characterize complex features,which considers the temporal and spatial patterns between trajectory locations,and integrates diverse analysis of trajectory environment.Moreover,the adaptive attention mechanism is also adopted in the paper to solve the influence of strong correlation between locations in the trajectory sequence,and improve the prediction accuracy.On this basis,for the implied correlation between users' moving trajectory under different spatial granularity,this paper adopts a hierarchical structure to excavate the hidden characteristics,so as to realize the effective capture and fusion of users' moving preferences,and further improve the prediction accuracy of the moving trajectory.In this paper,the proposed model is verified by experiments on two real datasets.The experimental results show that this method is feasible and effective for the prediction of users'moving trajectory.The results of this study are helpful to various location-based services and applications,and have certain reference value to the research of related work.