Long-term Truth Discovery Method based on GCN

Traditional truth discovery methods often face scalability challenges when processing large-scale dynamic traffic data. Moreover, traffic condition data exhibit significant spatial correlations and complex dependency structures, which traditional methods struggle to effectively capture. This limitation may lead to reduced accuracy in truth discovery. To address this issue, a long-term truth discovery method based on Graph Convolutional Networks (GCN) is proposed. Unlike conventional truth discovery approaches, this study models traffic data as a graph structure and utilizes GCN to process the graph, enabling the dynamic learning of latent relationships between source nodes. Additionally, by introducing a node reputation update mechanism, the method iteratively adjusts the reputation of each node based on its loss contribution, thereby modifying the weights of source nodes in subsequent training rounds. This iterative process enhances the reliability of truth discovery over the long term. Experimental results on real-world datasets demonstrate that the proposed method outperforms traditional approaches.