Personalized Recommendation Of Tourism Based On Social-Media

In the study of travel recommendation algorithms,the traditional content-based recommendation algorithms usually only consider the tourists preference,but ignore both the sentimental information of tourists and the theme of the attractions,thus affecting the effect of the recommendation.Based on the shortcomings mentioned above,this paper considers three factors,tourist preference,tourist attraction topic and subjective evaluation of tourist attraction,to promote the recommendation effect when designing the personalized recommendation algorithm of tourism.Finally,the validity of the modal algorithm is verified by building a data set by crawling the attraction and tourist information from the travel website.The work of this paper is as follows:A travel recommendation algorithm based on cross-domain topic sentiment model is proposed.When using the text information in the travel website for attraction recommendation,the characteristics of “polysemy” and “synonymy” in the text information are easy to cause semantic loss,and the degree of feature discrimination may be low if the sentimental factors are not considered.In view of the above two problems,this paper proposes a topic sentiment model to explore the potential topics and sentiments in the textual travel information and the coupling relationship between them.In addition,in order to better simulate the real decision-making process of tourism,this paper divides the data into user domain and attraction domain separately.Besides,user interests,topic similarities,and evaluation of attractions are considered in the recommendation algorithm to improve recommendation accuracy.The experimental results show that the proposed model has strong generalization ability and the recommendation algorithm has higher accuracy.A sentiemnt-aware recommendation algorithm for tourist attractions of multi-modal is proposed.When using the multi-modal data in the travel website for attraction recommendation,it is difficult to effectively integrate text,emotion and image information in the data.Aiming at this problem,this paper designs a sentiment_aware multi-modal topic model,which is a Bayesian three-layer model.By simultaneously modeling the three data of text,emotion and image,not only can the data distribution of a single modality be obtained separately,but also the relationship between different modalities can be learned at the same time.The proposed algorithm considers the important factors of “multi-modal data information”,“tourist's true preference”,“objective evaluation of attractions” and “tourism topic”.The experimental results show that the recommendation efficiency of the model is much higher than the traditional multi-modal recommendation algorithm that training by different modal data respectively and then re-synthesis scores.A text data set and a multi-modal data set for training the travel recommendation algorithm are constructed.Through web crawling technology,this paper crawled 14,648 tourist information and 8724 attractions information from the travel website named TripAdvisor.After preprocessing,the text dataset and multi-modal dataset of the user domain and the attraction domain are respectively constructed.Among them,the passenger information contains more than 450,000 text comments and more than 140,000 image data,and the scenic information contains more than 390,000 text comments and more than 80,000 image data.