Research And Application Of Railway Passenger Group Characteristics

The 14th five year plan points out that transportation is a basic,leading and strategic industry in the national economy.It is an important service industry and an important part of the modern economic system.It is an important support for building a new development pattern and a solid guarantee for serving the people's better life and promoting common prosperity.As its most important component,China's high-speed railway has completed the completion of "four vertical and four horizontal" passenger transport channels and the construction of "eight vertical and eight horizontal" road network,which provides powerful basic equipment for railway passengers to travel conveniently and quickly.However,the traditional extensive railway passenger transport marketing strategy can no longer meet the requirements of the increasingly competitive passenger transport market.In order to adapt to market changes,In recent years,the research of railway passenger transport operation management department mainly focuses on how to improve the utilization rate and service quality of passenger transport resources by adjusting passenger transport product design and optimizing passenger transport marketing strategy,so as to better meet the travel needs of passengers,so as to improve the competitiveness of passenger transport market.For example,by strengthening the optimization of train operation scheme and accurately implementing the research of "one map per day" to optimize railway passenger transport products;Provide expert support for passenger flow forecast by subdividing passenger flow by questionnaire;On the Beijing Shanghai line,the ticket price is dynamically adjusted based on the passenger seat rate of the train number to disperse the passenger flow and balance the passenger seat rate;Relying on the 12306 Internet Ticketing System of railway passenger transport,develop many service fields related to passenger travel,such as catering specialties,car hailing,hotels,passengers,air rail intermodal transport,highway rail intermodal transport and water rail intermodal transport,and provide one-stop service travel for passengers.The research and implementation of these strategies have contributed to the sustainable and healthy development of railway passenger transport,but there is a lack of analysis of the characteristics of passenger groups,There is no product design and marketing strategy optimization guided by passenger demand.Based on this,this paper uses big data and artificial intelligence technology to analyze passenger travel data,build passenger travel characteristics,social network relations,travel chain and other passenger digital feature systems,study the passenger group division model,so as to distinguish different groups of passengers,formulate personalized marketing and service strategies for different groups of passengers,and reasonably allocate railway passenger transport resources on the basis of passenger transport market segmentation,Further improve the allocation of transport capacity resources,alleviate the transport pressure of busy line trains,improve the revenue of railway extension service,realize the transformation of railway from traditional extensive marketing strategy to refined and even intelligent operation strategy,and expand it into a win-win strategy to maximize railway interests and optimize passenger service quality.The main research contents are as follows:(1)The digital characteristic system of railway passengers is constructed based on railway passenger identity information and travel information.(1)Based on the railway passenger transport marketing data,from the three perspectives of passenger characteristics,passenger transport product characteristics and the interaction behavior between passengers and products,the individual travel characteristics of passengers are constructed by using mathematical statistics,machine learning and other methods,including static characteristics such as gender,age and native place,dynamic characteristics such as the number of peers,travel frequency,high-end seat class and ride proportion,permanent residence,travel purpose,loyalty AI algorithm features such as value index;(2)Analyzing the travel data of railway passengers,66% of registered passengers have ticket purchase relationship,80% of passengers have peer relationship and 30% of passengers have integral transfer relationship.Based on this,combined with the travel characteristics of passengers,this paper constructs speculation strategies such as family relationship,work relationship and friend relationship in the real social network,and finally constructs a model to explore the potential relationship types in the passenger social network;(3)According to the travel time,departure city and arrival city,the passenger travel behavior constitutes a complete urban travel chain.Due to the limitations of railway passenger transport data,the hidden travel behavior of passenger railway is introduced to supplement the travel trajectory of passengers who cannot form a complete travel chain in data due to the choice of other means of transportation.Based on the analysis of passenger travel trajectories,passenger travel trajectories are divided into four categories;(4)Analyze the passenger railway hidden travel behavior,calculate the passenger railway travel loyalty,travel distance loyalty,travel od loyalty,etc.,calculate the section hidden travel proportion,so as to reflect the intensity of section competition,realize the competitive section identification based on passenger selection,and analyze the passenger characteristics of the competitive section Guangzhou Shanghai.(2)The segmentation model of railway passenger transport market is constructed based on the individual characteristics of passengers.At present,the individual travel characteristics of passengers designed by the portrait system have exceeded 5000.In order to overcome the problems of dimension disaster,over learning and local minimum points,the deep self coding neural network is introduced for feature selection,the lightgbm model is divided into sliding windows and optimized by depth,and the potential internal correlation information in passenger characteristics is deeply excavated by automatically subdividing the granularity of passenger feature analysis,So as to further improve the expression ability of passenger characteristics and realize the division of passenger groups.In view of the problems such as the deviation of the training results of the model caused by the limitations of the sample data in the actual production practice,the k-fold cross validation data set is used to divide the sample set into 8parts,and the passenger population division model is trained and tested respectively.The model fusion is best realized based on genetic algorithm,so as to improve the generalization ability of the model,In the same environment,the 2019 passenger travel data set is compared with multi classification twin vector machine,naive Bayesian classifier,k-nearest neighbor,random forest and other classification algorithms,and the results are evaluated.It is proved that the improved gap-rbf neural network has high performance under the same individual travel characteristics of passengers.(3)Build a segmentation model of railway passenger transport market integrated into passenger social network.Aiming at the problem that the existing passenger group division can not take into account the social relationship and individual characteristics,a passenger group division method integrating individual travel characteristics and social relations is proposed.The passenger personal travel characteristics are constructed according to the passenger travel data,and the topological maps of social relations such as family relationship,work relationship and friend relationship are extracted.The features of social relations are extracted through graph attention network and multi graph feature fusion,Finally,complete the division of passenger groups together with personal characteristics.In the same environment,the 2019 passenger travel data set is compared with the model and multi classification twin vector machine,and the results are evaluated.It is proved that the group division model integrating social relations has advantages over the model simply considering individual characteristics.(4)Build an adaptive passenger group division model for real-time stream processing to realize engineering high concurrency deployment.In order to improve the efficiency of the group classification model in passenger engineering deployment and real-time classification,the teacher student model is adopted,the group division model integrating social relations is taken as the teacher model with strong learning ability,and the knowledge learned is transferred to the student model with simple structure and weak learning ability,so as to enhance the generalization ability of the student model,and the compact model with less parameters is used to match the accuracy of the complex model,Realize the compression of the model.Based on knowledge distillation,use the complex teacher model to obtain knowledge in the training to guide the training of the smaller student model,make the student model imitate the teacher model,so as to realize the transmission of knowledge and the compression of the model,and improve the identification efficiency of the model while maintaining the accuracy of the group division model,so as to complete the engineering deployment of the passenger group division model.(5)Research on the application of railway passenger transport pricing strategy based on group division.There are differences in railway passengers and passenger transport products in different sections.In order to avoid the problems of overload of some trains and low occupancy rate due to uneven passenger flow distribution under a single ticket price,dynamically adjust the ticket price to guide the change of demand degree of different categories of passengers.In order to clarify the elastic demand of different groups of passengers for ticket prices,a price demand elasticity model based on passenger differences is designed.Different groups of passengers are divided through this model.On the premise of balanced distribution of railway sections,the functional relationship between price and passenger volume and elastic demand coefficient is established.The model is applied to Beijing Shanghai line to calculate the price elastic demand coefficient and its change law of different groups of passengers under price adjustment.(6)Research on railway extension service recommendation system based on group division.Railway passenger transport extension service products are attached to railway passenger transport products.Subdivide the passenger groups of different passenger transport products,analyze the preferences of different types of passengers to choose extension service products,and recommend personalized product lists for passengers.Based on the characteristics of passenger groups,this paper constructs the overall architecture of railway passenger travel service recommendation system,defines the main process of recommendation service,introduces product labels on the basis of passenger group division model,constructs recommendation algorithm,and solves the cold start problem by analyzing passengers' social network,so as to obtain the final recommendation result.