Research On Personalized Route Recommendation Model For Public Transportation

Due to the popularity of smartphones and mobile networks,travelers rely on map applications to plan their routes in advance.With the expansion of the city and the enrichment of public transportation,there are multiple paths for travelers to choose from between OD.As a reflection of the intelligence and humanity of the map application,the traveler’s preference for path selection can be obtained by analyzing the historical data of the traveler.Tired of making a choice from multiple paths,travelers prefer to be recommended a path to meet their needs during path planning.The path characteristics of public transportation and related researches in the field of personalized recommendation are combined to study the personal path selection model in urban public transportation.The main research contents are as follow:(1)Introduce and explain the blanks in the field of path personalized recommendation and related field research.The research framework of path personalized recommendation in urban public transportation is elaborated,which lay the foundation for the subsequent study of path selection models.(2)Design a questionnaire and analyze the results.Design a questionnaire for route selection in urban public transportation and collect basic information of travelers and route selection in different situations through the survey.The acceptable walking and riding distance intervals of travelers are obtained and the travelers’ ranking of the seven path features are learned.Through the selection results analysis of the paths selection questions,a conclusion can be reached that different travelers have different path selections under the same situation,which lay a realistic foundation for the research.(3)Establish a binary logit model of personal path selection.To study the sensitivity of travelers to the four path characteristics namely time,cost,walking distance,and number of stops,a binary logit model is built.The traveler’s path selection results in the questionnaire survey is divided into a training set and a test set.Regression analysis is performed in the training set to obtain the model parameters of each traveler.The obtained models can be verified in the test set and a relatively high accuracy is obtained.The feasibility and scientificity of the research are described.The conclusions drawn from a detailed analysis of the travelers’ parameter set can improve the accuracy of path recommendations during the cold start phase.(4)A factorization machine model for path selection is established.Based on the data collected by the questionnaire,the model parameters are solved and verified,proving that the factorization machine model,commonly used in the field of content and product recommendation,is suitable for solving path recommendation problems.(5)In order to learn the preference of traveler’s path selection more comprehensively,a comprehensive factorization machine model is proposed with a cross term with auxiliary coefficients introduced.The possibility of unpopular path selection is retained.By derivation,the model is proved linear complexity and parameters are multi-linear,which is suitable for existing parameter solving methods.Based on the questionnaire data,the model is solved and verified.The model can improve the novelty and accuracy of path recommendation,which proves the effectiveness and practicability of the model proposed in this paper.