A Social Network Detection Method Based on Graph Network Embedding

The exponential growth of social network information has a certain degree of falsehood, which seriously affects people’s judgment and decision-making and even leads to adverse consequences such as social unrest. In the research on intelligent detection of rumors, existing studies have overlooked the coupling relationship between nodes during the spread of rumors. Therefore, we propose a network comment relationship graph modeling method for social networks, which characterizes the interaction patterns and information dissemination mechanisms of social networks; A hybrid of graph convolutional neural network and graph multi-head attention mechanism is used to extract high-level semantic relationships between nodes during rumor propagation through bidirectional embedding expression, which can better perform feature extraction and graph network embedding in the comment relationship graph of rumor propagation. Based on the coupling characteristics of node similarity, importance, correlation, etc., a rumor detection model was designed through bidirectional feature fusion to improve the accuracy of rumor detection. The experimental results show that the model can better detect rumors on social networks. In comparison with the four benchmark models, the proposed model achieves an accuracy of 96.6% on the Weibo dataset, 85.5% on the Twitter15 datasets, and 91.8% on the Twitter16 datasets. The accuracy of the Weibo datasets and Twitter 16 datasets is higher than the benchmark model.