Research And Application Of Vehicle Travel Destination Recommendation Based On Node Embedding

As the popularity of the Internet,the data from people's behavior is becoming more and more accessible,and there is huge value hidden behind these data.Using these data,people's future behavior can be predicted.As for intelligent vehicles,the idea of future intelligent vehicles has gradually become a reality.Auto-driving,vehicle networking,human-computer interaction and location services are all hot research topics at present,and these function must be supported by behavior data.Intelligent vehicle is no longer a simple kind of transportation,but a high-tech product which is combined by different technologies in many fields such as vehicle engineering,computer science,internet technology,communication engineering and automation technology.Location service based on GPS(Global Positioning System)data perfectly matches the direction of intelligent vehicle development.Recommend destination for user by collecting,processing and analyzing travel data from intelligent vehicles not only bring user's pleasure,but also can help the government agent analyze the current urban traffic situation and make appropriate reminds,vehicle dispatching system can also judge the real-time road condition in time according to the forecast results,and plan more reasonable and efficient driving route,which has important significance for urban transportation and construction planning.Thus,by employing the GPS travel data from vehicles,we proposed a vehicle destination recommendation model based on location and user embedding,and we developed a relevant vehicle destination recommendation system.The main work of this paper is as follows:(1)The paper introduced the background,research significance and research status of the vehicle recommendation system.We introduced recommendation algorithm,machine learning algorithm,embedding technology and travel data preprocessing technology based on geographic location.(2)We analyze the data of vehicle traveling,and this paper explore and summarize the correlation between time and destination recommendation.Then,we analyze the travel behavior of user and meaning of geographical grid.(3)We proposed a vehicle destination recommendation model based on location and user embedding by using vehicle historical travel data.Firstly,the GPS travel data from vehicle is preprocessed.According to the longitude and latitude of the vehicle's source and destination,the geographic grid technology is used to discretize into geographic grid label.Thus,the recommendation of longitude and latitude is transferred into the recommendation of geographic grid.Second,using the model based on node embedding to make vehicle destination recommendation.According to the historical record of the vehicle,the Bayesian model is used to maximize the posterior probability of the destination,and employ location embedding to mine the relationship between source and destination to further improve the prediction accuracy.In the experiment,the experimental results of the proposed algorithm are compared with other algorithms.(4)According to the proposed model,we developed a vehicle destination recommendation prototype system,which includes historical travel information injection,vehicle destination recommendation,POI recommendation and so on.The user can inject the user's travel information by using the injection function,and the destination recommendation function predicts destination for user.