Research On User Profile Modeling And Personalized Recommendation Based On Tourism Big Data

With the continuous development of the tourism market and the upgrading of tourism user needs,how to provide personalized,diversified,and high-quality tourism recommendation services for users has become an urgent problem to be solved.User profiles based on user data analysis can fully reflect the characteristics and personalized needs of users,making personalized recommendations based on user profiles gradually widely used in fields such as e-commerce,library and information,and tourism management.However,existing research mostly focuses on unilaterally characterizing user features when constructing user profiles.The dimensions of constructing profiles are relatively single,resulting in incomplete characterization of user features and demand preferences,which affects recommendation effectiveness.Therefore,multidimensional and meticulous characterization of user features has become the focus of research.On the basis of collecting user information on the Mafengwo tourism platform,this article proposes to construct a multidimensional user profile model,and then combines recommendation technology to provide personalized recommendations to users.The main research content is as follows:(1)Build a multi-dimensional user profile model.By mining users’ travel notes and extracting their specific contextual labels,a contextual dimension portrait of users is constructed to analyze their contextual characteristics and preferences.The potential Dirichlet topic model is used to extract the topics and interest feature words that users pay attention to from the travel notes data generated by users,and build the user’s interest profile.Use natural language processing technology to extract attribute words emotional word pairs from users’ comments on scenic spots,and conduct emotional rating and emotional polarity judgment on them.Finally,extract emotional polarity tags from each attribute surface to construct user emotional dimension portraits.By using a user profile model to depict users’ contextual,interest,and emotional features,this lays the foundation for personalized recommendation methods based on user profile models.(2)Personalized recommendation based on user profile model.Recommend association rules based on user context dimension profiling.Analyze the correlation between the user context preferences and contextual features reflected in the user context profile and the scenic spots selected by the user,and then match the target user’s contextual features and preferences with the rule context in the association rule library to extract rules with high similarity.Based on the scenic spots after the rules,recommend the target user to obtain the recommended item set A.Recommend similar users by combining user interest profile and emotional profile.Calculate the similarity of interests between users based on their interest feature words reflected in the user interest profile.Based on the emotional polarity and rating of users towards scenic spots in various attribute planes reflected by the user’s emotional dimension profile,a user attraction attribute emotional score matrix is constructed,and then the emotional similarity between users is calculated.By combining interest similarity and emotional similarity,find "nearest neighbor users" that are similar to the target user,and obtain the recommended item set B through similar users.Finally,the scenic spots recommended based on association rules and similar user recommendations will be integrated to obtain a mixed recommendation project set C,which will be recommended to users.(3)Verification of personalized recommendation effectiveness based on user profile model.In this paper,we collected the data set of Mafengwo tourism platform for empirical analysis,combined with the evaluation index of recommendation effect,and compared the recommendation effect of the recommendation method in this paper with the traditional collaborative filtering recommendation method and the recommendation algorithm based on similar users.The results indicate that the personalized recommendation method based on user profile model in this article has good recommendation performance,and to some extent,it improves the shortcomings of traditional recommendation methods due to data sparsity,cold start,and other reasons,resulting in lower recommendation quality.