Research On Personalized Recommendation Method Of Tourist Scenic Spots Based On Knowledge Graph

With the development of information technology and the popularity of the Internet,the data on the network has grown exponentially and the problem of information overload has become a major problem for users to obtain information.In order to discover user's interest from mass information and meet personalized information needs of users,recommendation system came into being.The traditional collaborative filtering recommendation mostly uses rating data to recommend,which has the problem of data sparsity and limit the performance of the recommendation system.There is a wealth of knowledge in knowledge graph,which will be useful auxiliary data in recommendation system.Therefore,the recommendation based on knowledge graph has attracted more and more researchers' attention.The personalized scenic spot recommendation method based on knowledge graph is studied in this thesis.By constructing the knowledge graph of scenic spots,extracting the entity vectors,the user interest model is constructed which then was integrated into the matrix decomposition recommendation model,so as to improve the performance of the scenic spot recommendation system.The main contents of this thesis are as follows:(1)A scenic spot knowledge graph based on ontology was constructed.At present,there is no public scenic spot knowledge graph in the Chinese field,which needs to be built according to the needs.By analyzing the construction method of domain knowledge graph,an ontology-based scenic spot knowledge graph was designed and implemented.Firstly,according to the purpose of ontology,reusable ontology was looked for,concepts and classifications were designed,the attributes of classes were defined,and scenic spot ontology was constructed.Then,data sources were collected,and scenic spot knowledge was extracted based on ontology.A string-based and structure-based entity alignment method was designed,which aligns entities from two levels of string and entity attributes,respectively.Finally,the scenic spot knowledge graph was stored in Neo4 j graph database.(2)A user interest vectorization method based on knowledge graph was designed.The common user interest modeling methods were analyzed,which are not suitable for user interest modeling based on knowledge graph.By learning the representation of knowledge graph,the knowledge in knowledge graph was vectorized.On this basis,a user interest vectorization method based on attributes of entities in knowledge graph was proposed.By analyzing the attributes of scenic spots,it is found that there is an association between the attributes of scenic spots and users' interests.Based on the vectorization of knowledge graph,the weight function of the attribute of scenic spots was defined.Combined with the attribute value of historical visited scenic spots,the user interest vector was constructed,which can accurately locate the user's interest in the attribute of scenic spots and be used for subsequent scenic spots recommendation.(3)A matrix decomposition model integrating knowledge graph was proposed.By analyzing the collaborative filtering recommendation method,rating data was used only in the traditional collaborative filtering method for recommendation,which has the problem of data sparsity.On the basis of user interest vectorization,user preference similarity was calculated and user implicit feedback was jointly used to perform matrix decomposition so that knowledge graph information could be integrated into matrix decomposition model,which makes up for the shortage of matrix decomposition algorithm that does not consider the attribute information of scenic spots and alleviates the problem of data sparsity.The results of comparison experiments show that the proposed method of personalized scenic spots based on knowledge graph is effective.Compared with the baseline recommendation method,the precision and the recall of recommendation have been improved to a certain extent.