Real-Time Emotional Topic Recommendation in Social Media News Using MDT and Hypergraph-Based Neural Networks

Abstract

In the intelligent recommendation of user emotional topics within social media news dissemination, the system must adapt to the dynamic changes in user interests and enhance the real-time accuracy of recommended content. To address this, a multiple decision tree (MDT) recommendation model is first proposed, which combines with a neural network structure optimization model based on hypergraphs (NNSO-hy) to construct a user interest topic recommendation model. Additionally, to cope with the rapid change of news topics, dynamic window technology is used to capture the dynamic changes of news topics and to adjust the recommended content, where the short text clustering model is designed to optimize the objective function, utilizing an improved Non-negative Matrix Factorization (NMF) model to cluster the window news. The experimental results show that the proposed model effectively captures news topics in real-time across different datasets. Moreover, our precision rates for dynamic burst topic detection are 0.72, 0.85, and 0.83, respectively. When conducting intelligent user sentiment topic push notifications on the three datasets, the precision rates are 0.74, 0.85, and 0.84, respectively. After 30 iterations, the loss function of MDT-NNSO-hy stabilizes, achieving an accuracy (P), recall (R), and F1 score of 0.82, 0.81, and 0.87, respectively, on average across three datasets. In summary, this study realizes the intelligent push of users' emotional topics in social media news dissemination by integrating MDT, NNSO-hy, and dynamic topic detection technology. This not only improves the accuracy and real-time performance of recommendations but also provides new ideas and methods for the development of personalized recommendation technology.