Rresearch On Attractions Recommendation Method Based On Graph Neural Network

Under the guidance of national policies and the promotion of industry demand,the tourism market develops rapidly,and the problem of information overload appears.It is an urgent problem to quickly and accurately select the tourism resources that users are interested in from the overloaded information.Based on the above background,tourism recommendation system came into being.The current tourism recommendation methods are divided into two categories:traditional recommendation and deep learning recommendation.Due to the problems of low tourism frequency and sparse interactive data,the current common recommendation algorithms,such as collaborative filtering and matrix decomposition,are not effective.Before deciding on tourism activities,users are used to learning about tourism resources information from the information overloaded Internet.There are many kinds of tourism information in the current tourism websites,and they mainly display tourism resources information,without the function of providing personalized scenic spots,routes and other tourism resources for users.With the emergence of recommendation system users can accurately select the resources they are interested in from the massive tourism information.The travel recommendation system first analyzes the needs of users,and then obtains the most relevant resources through related technologies to assist users' travel.In view of this,this paper formalizes the behavior of users visiting scenic spots into heterogeneous graph structure,designs a scenic spot recommendation algorithm based on graph convolution neural network,and realizes the personalized tourism recommendation system.The specific research contents are as follows:First of all,users and scenic spots are formalized as characteristic points,so users and the scenic spots they visited naturally form a heterogeneous graph,so the task of recommending scenic spots for users is transformed into the problem of link prediction of nodes in the heterogeneous graph.On this basis,this paper designs a scenic spot recommendation algorithm based on graph convolution neural network,which aims to learn the embedded representation of users and scenic spots by using multi-layer graph convolution operation to explicitly model high-order connectivity.The updated embedded representation of users and scenic spots is input into multi-layer perceptron,and the nonlinear interaction between users and scenic spots is captured by multi-layer perceptron,In this way,we can capture the cooperation signal hidden in the user's interactive information of scenic spots,and ultimately improve the recommendation effect.Secondly,the algorithm of scenic spot recommendation based on graph neural network is verified by experiments.The parameters in the experiment were analyzed,and the number of layers of the graph convolutional neural network and the length of the recommendation list were taken as examples to test the variation of the algorithm recommendation performance under different parameters.Analyzed the relationship between the convergence of the algorithm and the number of iterations and the variation of the hit rate and normailized discounted cumulative gain of the recommended algorithm is analyzed under different number of iterations.The performance of this recommendation algorithm is analyzed and compared with the relevant recommendation methods of scenic spots.The experimental results show that compared with other algorithms,the recommendation algorithm in this experiment improves the hit rate and the normailized discounted cumulative gain of the algorithm to a large extent,which indicates the effect of the recommendation algorithm.Finally,the personalized tourism recommendation system is designed and implemented.System simulation method is used to analyze the performance of the system,and the requirements of the system are analyzed from the functional and non-functional aspects.Design the system architecture,system function modules and main database tables.The usability of the algorithm is verified by the realization of the attractions recommendation module,the trip planning module and the personal homepage.Test the development environment of the system,and test the main functional modules of the system.