Personalized Route Navigation And Travel Time Prediction Based On Vehicle Trajectories

As an important component of intelligent transportation,route navigation and planning play an important role in our daily life.In recent years,the "non-discriminatory" navigation services and frequent traffic jams have caused more concern to the research on how to improve the quality of navigation services.The study of personalized navigation and travel time prediction in this paper has importantly theoretical and practical value.The key issue of personalized navigation is user preference modeling.Most of the existing methods use preference vectors to represent user preferences,the method of getting the preference vector has linear fitting and distribution fitting.Based on method of the Gaussian mixture distribution model of user preferences,an improved method of personalized recommendation has been put forward in this paper.It introduces the steering overhead and describes the user's preferences in more detail;In order to reduce the impact of accidental factors,this paper also introduces a frequency filtering mechanism.Compared with the comparison method,the recommendation effect is improved.Accurate travel time prediction is critical for navigation routes.This paper improved KNN prediction algorithm which is based on similar trajectory.When retrieving similar track,it not only considers completely consistent trajectory at the starting and ending point,also considers the path through the starting and ending point.At the same time,in order to improve the prediction accuracy of the algorithm,the other external factors are also considered,in addition to the time sequence characteristics.The premise of KNN prediction algorithm is the rich history trajectory,therefore,to solve the problem of "cold start" and arbitrary path prediction problems in complex urban road network,this paper also proposes a prediction method which is on the basis of feature fusion of ensemble learning.Besides general characteristics of stroke,this paper also considers the number of road intersection and road turn.Finally,with ensemble learning model predicting the road in urban city,results show improved prediction accuracy.After a series of data preprocessing on Cheng Du trajectory data,such as trajectory restoration,trajectory segmentation,abnormal point removal and map matching,multiple experiments were designed to verify the effectiveness of the above methods.Firstly,the improved personalized navigation method is compared with comparison method and the results show that the improved method is better than the comparison method.Then,the improvement effect based on KNN prediction algorithm is verified,and the retrieval effect and prediction accuracy are improved.Finally,compared with the prediction method based on the road section,sub-path,the conclusion shows that the prediction accuracy of the ensemble learning method is superior to the comparison method.