LOSS-GAT: Label propagation and one-class semi-supervised graph attention network for fake news detection

Abstract

In today’s world of social networks, fake news spreads quickly and causes serious problems. This has made it crucial to develop automated systems to detect and combat disinformation. Machine learning and deep learning are often used to identify fake news, but they struggle due to the lack of labeled news datasets. To address this, the One-Class Learning (OCL) approach uses a small set of labeled data. On the other hand, representing data as a graph enables access to diverse content and structural information, and label propagation methods on graphs can be effective in predicting node labels. In this paper, we adopt a graph-based model for data representation and introduce a semi-supervised and one-class approach for fake news detection, called LOSS-GAT. Initially, we employ a two-step label propagation algorithm, utilizing Graph Neural Networks (GNNs) as an initial classifier to categorize news into two groups: interest (fake) and non-interest (real). Subsequently, we enhance the graph structure using structural augmentation techniques. Ultimately, we predict the final labels for all unlabeled data using a GNN that induces randomness within the local neighborhood of nodes through the aggregation function. We evaluate our proposed method on six common datasets and compare the results against a set of baseline models, including both OCL and binary labeled models. The results demonstrate that LOSS-GAT achieves a significant improvement in performance, with enhancements ranging from 5% (on the FEVER dataset) to 20% (on the FakeNewsNet dataset) in terms of the Macro-F1 metric, all while utilizing only a limited set of labeled fake news data. Noteworthy, LOSS-GAT even outperforms binary labeled models.