Implicit relationship multi granularity coupled graph neural network algorithm for new media content recommendation

Abstract

To alleviate information overload and improve the accuracy and personalization of new media content recommendations, a hidden relationship, multi-granularity, coupled graph neural network algorithm are proposed for new media content recommendations. Based on an in-depth exploration of the implicit relationship between users and content, a three-layer coupling graph mechanism of user background content is constructed. Combined with the powerful representation ability of graph neural networks, representation learning is carried out from two perspectives: intrinsic and extrinsic motivation. The experiment shows that compared with the recommendation algorithm without implicit relationship calculation, the mean square error of this algorithm is reduced by 43.90 %. Meanwhile, the accuracy and recall of the algorithm are 94.46 % and 93.58 %, respectively, and the recommendation efficiency is improved by 41.67 %. Moreover, when faced with different content recommendation requirements, its accuracy and recall rate always remains above 90 %, and it also shows better stability and accuracy when dealing with sparse data and cold start problems. The results indicate that the proposed implicit relationship multi granularity coupled graph neural network algorithm can effectively improve the quality and efficiency of new media content recommendation. It can provide a practical and feasible solution to solve the problem of information overload in new media and meet users’ personalized needs.