Travel Planning Model And Application Of Air-Rail Intergrated Service Based On User Portrait

With the continuous extension and expansion of China’s air and railway network,China’s aviation and railway mileage has grown very steadily,and it has become the two main choices for residents’ daily mid-and long-distance travel.The two have been a longterm competitive relationship.With the introduction and development of Maa S(Mobility as a Service)and the premise that the railway network has almost penetrated into most cities in my country,the combined travel of aviation and railway has also become a popular choice.However,on the current ticketing platform,there is a lack of design for air-rail joint travel,and there is a lack of research on the push of personalized solutions for air-rail joint travel.The joint ticket recommendation of the ticketing platform is generally difficult to find and the transit city is limited.The transfer time is unreasonable and the number of plans is small.In this context,this article studies the generation of connecting plans for different types of passengers.This article relies on the National Key Research and Development Program(2018YFB1601300),the National Natural Science Foundation of China(52072066),the Jiangsu Outstanding Youth Fund Project(BK20200014),and the Jiangsu Transportation Science and Technology Plan Project(2020Y12).This article first reviews user portraits and multi-modal travel plans,and determines the research methods.Secondly,this paper cleans and expands Ctrip’s real air-rail joint travel data,extracts travel characteristics,and visualizes the distribution of transfer locations,transfer trajectories,transfer time characteristics,etc.,and uses K-means clustering.The method classifies the groups of people on the same route,constructs group user portraits,divides the groups into cost-sensitive,time-sensitive,and intermediate passengers,and compares the differences in travel behavior between the classes.The clustering effect is good.Based on the classification of people,this article considers travel cost and travel time as the objective function,and establishes a multi-objective planning model.In order to reflect the differences in passenger group preferences,the weight coefficient is introduced to process the objective function,and the DQN algorithm is designed to model On the basis of proving the adaptability of the algorithm to the problem in this paper,the reward and punishment learning mechanism of reinforcement learning is used to combine the time and cost with the reward and punishment value,design the algorithm framework,clarify the algorithm flow,and achieve the maximum objective function after multiple iterations.Optimize and solve to get the set of connected programs.Finally,taking Beijing-Guangzhou as an example,the transit city set and transit time are calibrated according to the actual travel data,and the DQN algorithm is used for three different groups of people to obtain a set of differentiated and personalized connection schemes.At the same time,the algorithm obtained Comparing the plan set with the air-rail joint plan set of similar ticketing platforms,it is found that it has the advantages of reasonable transit time,controllable transit plan collection,consideration of crowd preferences,strong pertinence,and large selection of transit locations.