Detection and optimization of short video information splitting using graph neural networks

作者：Meng Qiu, Tingting Wang · 年份：2025 · DOI：10.1117/12.3093310 · 研究领域：Advanced Data and IoT Technologies、Big Data and Digital Economy、Advanced Graph Neural Networks

This paper proposes a short video information splitting detection method based on graph neural network. By constructing a heterogeneous graph of user-video interaction, graph convolution and attention mechanism are used to achieve accurate identification of split content. The designed SplitGNN model captures the topological characteristics of the propagation network through multilayer feature aggregation, and optimizes the detection performance by combining temporal dynamic modeling. Experimental results show that the model is significantly better than traditional methods in accuracy, precision and recall, with an accuracy of 0.90 and an F1 score of 0.88. The attention mechanism enhances the ability to extract key interactive features, while temporal dynamics improves the model's adaptability to propagation changes. The study verifies the effectiveness of graph neural network in dealing with the problem of short video information splitting and provides technical support for platform content governance.