| With the explosive growth of tourist information and complex diversity of specific needs,in order to find suitable tourist attractions from a vast amount of information,tourists need to consider factors such as hobbies,time,prices,location and so on.The search process will consume considerable amounts of energy and money,and also reduce the user's travel experience.In order to solve this problem,personalized recommendations for tourist attractions came into being.It is of good practical value to conduct accurate and efficient personalized recommendation research of tourist attractions.The traditional personalized recommendation algorithm usually uses user ratings or historical tourist attractions to represent interest preferences and user similarity,ignoring the user's needs at different times and the characteristics of the attractions themselves,and the accuracy of similarity calculation is low,the recommendation of new attractions is bad.Although the tag can be used to express the characteristics of the scenic spot to a certain extent,the user prefers to mark the resources freely according to personal preferences,which results in the marked coverage of the tags being small and arbitrarily large and one-sided.Moreover,fewer tags about attractions on the travel websites,which lack uniform classification standards.To solve the above issues,this paper does the following work:(1)Enrich tags.Firstly,this paper comprehensively integrates the visual perception points and perceived image classification criteria of tourist destinations,and formulates the attraction feature table that meets the needs of this article,and considering the characteristics of Qingdao and attractions type tags published by ctrip,supplemented the feature table,as the standard classification standard for attractions;then,it extracts the keywords that match the attraction features from the scenic spot reviews and classifies them in the corresponding attraction feature tags;finally,use site-supplied attraction type tags to enrich the extracted tags,taking into account the features of the spot itself.(2)Improve the calculation of similarity.This paper proposes to use spot feature tags instead of scenic spots itself,and to describe the user's interest preferences based on the feature tags of user-preferred spots,and to solve the similarity between users,to alleviate the cold start problem of new users to some extent,and to effectively distinguish the same types of spots.Moreover,similarity thresholds are used to solve similar neighborhood users of target users,which effectively avoids the situation of using less similar users as neighbor users,and improves the accuracy of similar neighbor users.(3)Improve personalized attractions recommendation algorithm.Firstly,the tags associated with attractions,events,time,attraction tickets,cultural customs,entertainment facilities,shopping,gourmet snacks and other related factors are combined to enrich the tags that can be used to characterize attractions and interest preferences;then,it is combined with the traditional recommendation technology,and the recommendation results of the tag-based recommendation algorithm and the tag-based collaborative filtering algorithms are respectively analyzed;finally,the corresponding weights are given to the recommended results of these two algorithms,and they are simply summed,effectively combining the advantages of the two recommended algorithms.The recommendation results of the tag-based hybrid recommendation algorithm are analyzed,the accuracy and recall rate of the recommendation are improved,the recommendation effect is good,and the user's travel experience is enhanced. |
