Scholarlink

<p>Nowadays, the expansion of online social networks resulting in the spread of social news. Some social networks including Twitter, Instagram and Facebook have been impacted by user-generated fake news. Unfortunately, some individuals utilize unethical tactics to increase their links and reputation. Recent information on online social networks is often inaccurate and can mislead other users. Fake news is purposely propagated to deceive readers, making it harder to detect through shared information. This research proposes EnsembleNet: A Novel Deep Ensemble Learning Model with GANBERT and BiLSTM for Automated Fake News Detection (DEL-BiLSTM-AFND) to tackle this problem. Initially, the input data are gathered from ISOT Fake news dataset and Politi Fact dataset, followed by preprocessing for tokenization. The Generative Adversarial Networks-Bidirectional Encoder Representations from Transformers (GANBERT) is used to generate high quality word embeddings from the news articles. Then the Convolutional Neural Network is used to extract key textual features from word embeddings. Then the Bidirectional Long Short-Term Memory is considered to classify news articles fake or real depending upon the sequential dependencies. At last, the performance of proposed DEL-BILSTM-AFND approach offers 20.35, 24.62 and 21.56% higher accuracy and 23.10, 21.19 and 24.21% higher f1- score compared with existing models: content-dependent transfer learning method for fake news detect in social media (TLA-FND), Word embedding and categorization models and their effects on fake news detection (WEC-FND) and Modeling Context and Content Features for Fake News Detection (MCCF-FND) methods, respectively.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

EnsembleNet: a novel deep ensemble learning model with GANBERT and BiLSTM for automated fake news detection

  • 期刊论文

  • 发表日期：

    2025年8月22日

查看全文
  • B. M. Brinda,
  • C. Rajan,
  • K. Geetha

摘要

Nowadays, the expansion of online social networks resulting in the spread of social news. Some social networks including Twitter, Instagram and Facebook have been impacted by user-generated fake news. Unfortunately, some individuals utilize unethical tactics to increase their links and reputation. Recent information on online social networks is often inaccurate and can mislead other users. Fake news is purposely propagated to deceive readers, making it harder to detect through shared information. This research proposes EnsembleNet: A Novel Deep Ensemble Learning Model with GANBERT and BiLSTM for Automated Fake News Detection (DEL-BiLSTM-AFND) to tackle this problem. Initially, the input data are gathered from ISOT Fake news dataset and Politi Fact dataset, followed by preprocessing for tokenization. The Generative Adversarial Networks-Bidirectional Encoder Representations from Transformers (GANBERT) is used to generate high quality word embeddings from the news articles. Then the Convolutional Neural Network is used to extract key textual features from word embeddings. Then the Bidirectional Long Short-Term Memory is considered to classify news articles fake or real depending upon the sequential dependencies. At last, the performance of proposed DEL-BILSTM-AFND approach offers 20.35, 24.62 and 21.56% higher accuracy and 23.10, 21.19 and 24.21% higher f1- score compared with existing models: content-dependent transfer learning method for fake news detect in social media (TLA-FND), Word embedding and categorization models and their effects on fake news detection (WEC-FND) and Modeling Context and Content Features for Fake News Detection (MCCF-FND) methods, respectively.