Research On Collaborative Filtering Recommendation Of Fresh Agricultural Products Based On Graph Neural Networks

With the development of information technology,online purchasing of fresh agricultural products has become an important consumption method.The data of fresh agricultural products presented on the network is diverse and complex.It is a valuable research topic to find fresh agricultural product information that meets user needs from a large amount of information.Collaborative filtering recommendation,as a traditional and main recommendation method,has been widely used in various e-commerce platforms in recent years.However,most existing collaborative filtering recommendations only consider the interaction behavior between users and products,ignoring users’ emotional preferences,and the existing collaborative filtering algorithms are difficult to handle graph-structured data,which affects the recommendation effect.This thesis mainly studies the collaborative filtering recommendation of fresh agricultural products based on graph neural networks and proposes a graph neural network-based collaborative filtering recommendation model for fresh agricultural products.The specific work of this thesis is as follows:(1)In response to the problem that collaborative filtering recommendation models rarely consider the specific features of fresh agricultural products and users’ emotional preferences,a user-sentiment label-fresh agricultural product association graph is constructed.First,Text Rank algorithm is used to extract keywords from users’ comment texts that can reflect the specific features of fresh agricultural products,and the improved ATAE-LSTM model is used to extract the sentiment polarity of users’ comments on these keywords,which are regarded as sentiment labels containing users’ sentiment polarity.Finally,the user-sentiment label-fresh agricultural product association graph is constructed by combining user ratings.(2)Based on the above association graph,a graph neural network-based collaborative filtering recommendation model for fresh agricultural products is proposed.The association graph is divided into emotion preference association graph and user-fresh agricultural product interaction graph,and collaborative filtering and graph neural network technology are used to obtain the embedding representation of each user and fresh agricultural product in each graph,respectively.The embedding representation of users and fresh agricultural products in different graphs are then fused to obtain the final user and fresh agricultural product embedding representation,and the target user is recommended.Experiments on a real dataset crawled from the JD e-commerce platform verified that the proposed model has certain advantages in terms of Hit Rate(HR)and Normalized Discounted Cumulative Gain(NDCG)and other indicators.