Research On Attractions Recommendation Method Integrating Knowledge Graph And User Interest Model

The exponential growth of information and knowledge on the web in recent years has created the problem of "information overload" for users.Through various behavioural data of users,it is possible to tap into their interests and thus personalise recommendations.The most common type of user interest data is rating data,but the disadvantage of rating data is that it cannot present all of the user’s interests and has the problem of sparse data,which affects the effectiveness of recommendations.Research shows that users’ choice of items is actually a choice of different attributes of items,so the study of item attributes can help to better understand users’ preferences.In this paper,we choose to introduce knowledge graphs into the recommendation method,on the one hand as auxiliary data that can alleviate some of the problems that existed in previous collaborative filtering.On the other hand,through the effective use of items and their attributes,we can better tap into users’ personalised interests and thus achieve more personalised yet accurate recommendations.The research in this paper is divided into the following three main areas.（1）Building a knowledge map of the Australian tourism domain based on domain knowledge modelling.Firstly,a conceptual hierarchy model of the Australian tourism domain is constructed,and then data is collected and processed to achieve knowledge acquisition and obtain a knowledge triad.Next,the entities obtained from multiple knowledge bases are aligned and knowledge is fused.Finally,the resulting data is stored in a Neo4j database.（2）Mining user preferences for attractions and constructing user interest models.By converting users’ preferences for attractions into preferences for attributes of attractions,the knowledge representation is used to vectorise the knowledge graph to obtain the attribute vector of attractions,so as to calculate the user’s visit interest vector.（3）A hybrid recommendation method that integrates the knowledge graph of attractions and users’ interest in visiting them is proposed.On the one hand,the similarity of user preferences is calculated using the knowledge graph-based user interest model,and the predicted ratings of attractions are obtained using matrix decomposition methods.On the other hand,what we should do is mining the semantic information in the knowledge graph and combining the user’s historical rating data.The final recommendation list is obtained by merging the recommendation lists obtained after ranking the predicted scores from the two aspects.This paper proposes a method for recommending attractions that incorporates knowledge graphs and user interest models.The experimental validation results against other comparative algorithms reveal that the algorithm proposed in this paper is optimised in terms of recommendation performance.