Rumors detection in social networks using dynamic graph-structured bi-directional long-short term memory technique

The rumor detection, several algorithms retrieve rumor attributes via textual semantics and the rumor's transmission structure. However, most of these methods fail to account for the fact that erroneous and unnecessary contacts in the rumor's transmission pattern would lower the detection rate. In addition, several previous rumor detection algorithms were not able to successfully extract crucial indications from the responses of social network members. In light of these occurrences, this article presents a unique Dynamic Graph-Structured Bi-Directional Long-Short Term Memory (DG-BiLSTM) techniques leverages dynamic graph structures, bidirectional temporal processing, and real-time filtering mechanisms to effectively handle erroneous and unnecessary contacts in the rumor's transmission pattern, ultimately improving the accuracy of rumor identification in social networks.At first; we get data samples from actual publicWeiboand twitter datasets. The preparation stage should normalize the raw data because the texts in the raw data may contain duplicate data. To extract the interactive semantic characteristics from the normalized data, two approaches are used i.e. Term Frequency-Inverse Document Frequency (TF-IDF) and Bag of Words (BoW). Then, new component subsets are created from the extracted data using the Genetic algorithm. The suggested method is then utilized to extract the rumor texts from the chosen data. Additionally, the precision, f1-score, recall, and accuracy based on the proposed method's performance are assessed and contrasted with those of existing approaches. The proposed system performed exceptionally well in the rumor detection challenge, demonstrating its viability and efficacy. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.