Research On User Travel Route Recommendation Method

With the rapid development of urbanization and the continuous growth in vehicle ownership,the traffic condition in the city is becoming more and more complicated.Using the improved weighting algorithm to calculate the fastest or shortest route,traditional route planning methods usually recommend a congested path and dissatisfy user without considering the actual traffic situation and the travel preferences of the user.Therefore,the topic of how to recommend an effective route for travelers is always a hot issue for researchers.With the popularity of GPS devices,people can easily obtain historical travel trajectory data of users.The historical track data contains the user's travel preferences.Therefore,trajectory-based route recommendation methods have gradually attracted people's attention.Existing leading trajectory-based path planning algorithms heavily rely on the quality of historical trajectories.This kind of algorithm performs poorly when uneven trajectory distribution results in sparse trajectories in some regions.Aiming at the problem that existing path planning methods can not satisfy the travel demand of the user,this paper proposes a path planning model named 2P++,which considers both the path length and travel preferences of the user.First,inspired by LSTM,a new method extracting travel preference features from trajectory data was proposed to learn the long-term dependencies between nodes in the path.Secondly,the MCMC sampling technology is introduced into the A* algorithm,and a named dual-target path search framework is proposed.In addition,in order to increase the efficiency of the algorithm when searching for too many nodes in the large-scale graph data query path in the 2P++ algorithm,this paper optimizes the path query process and uses the graph attention network to estimate the distance between nodes to improve the evaluation function in the path planning algorithm.This method reduces the number of extended nodes in the path search process and speeds up the query path activity.In the end,the author uses the taxi trajectory and electronic map of Beijing forexperimental analysis.The final experimental results show that comparing with the traditional shortest path planning algorithm and the leading trajectory-based path planning algorithm,the model achieves better results on recommended path length,accuracy,and travel time.