Scholarlink

<p>With the rapid growth of social media, the spread of fake news has become a serious challenge with far-reaching consequences for global societies and policies. This paper proposes an innovative hybrid method for fake news detection, in which a stacked ensemble classifier—comprising support vector machine (SVM), random forest (RF), logistic regression, KNN, gradient boosting, and Naïve–Bayes—is optimized using a genetic algorithm and empowered by a novel hybrid feature extraction approach based on the BERT model. The proposed framework leverages the deep semantic capabilities of BERT embeddings to generate highly discriminative feature vectors and integrates multiple diverse machine learning algorithms into a unified stacked structure, whose weights and parameters are fine-tuned via a genetic algorithm. To address data imbalance, generative adversarial networks (GANs) are employed to create realistic synthetic samples, ensuring balanced representation of classes. Experimental evaluation on two standard datasets—LIAR and ISOT—demonstrates the superior performance of the proposed method, achieving 98.05% accuracy on LIAR and 99.1% accuracy on ISOT, with consistently high scores in precision, recall, and F1-score. These results confirm that the synergy between stacked ensemble architecture, GA-based optimization, and BERT-powered feature extraction is the primary innovation driving the method’s success over existing approaches. Furthermore, due to the massive volume and velocity of social media streams, our framework is designed to leverage high-performance computing (HPC) infrastructures and parallel/distributed processing environments. This enables scalable and real-time fake news detection while maintaining computational efficiency.</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

A hybrid method based on optimized ensemble classifier using genetic algorithm and novel embedded feature extraction based on BERT for detecting fake news in social media

  • 期刊论文

  • 发表日期：

    2025年9月28日

查看全文
  • Kareem Awad Dawood,
  • Golnaz Aghaee Ghazvini,
  • Fariba Majidi,
  • Ali Albu-Rghaif

摘要

With the rapid growth of social media, the spread of fake news has become a serious challenge with far-reaching consequences for global societies and policies. This paper proposes an innovative hybrid method for fake news detection, in which a stacked ensemble classifier—comprising support vector machine (SVM), random forest (RF), logistic regression, KNN, gradient boosting, and Naïve–Bayes—is optimized using a genetic algorithm and empowered by a novel hybrid feature extraction approach based on the BERT model. The proposed framework leverages the deep semantic capabilities of BERT embeddings to generate highly discriminative feature vectors and integrates multiple diverse machine learning algorithms into a unified stacked structure, whose weights and parameters are fine-tuned via a genetic algorithm. To address data imbalance, generative adversarial networks (GANs) are employed to create realistic synthetic samples, ensuring balanced representation of classes. Experimental evaluation on two standard datasets—LIAR and ISOT—demonstrates the superior performance of the proposed method, achieving 98.05% accuracy on LIAR and 99.1% accuracy on ISOT, with consistently high scores in precision, recall, and F1-score. These results confirm that the synergy between stacked ensemble architecture, GA-based optimization, and BERT-powered feature extraction is the primary innovation driving the method’s success over existing approaches. Furthermore, due to the massive volume and velocity of social media streams, our framework is designed to leverage high-performance computing (HPC) infrastructures and parallel/distributed processing environments. This enables scalable and real-time fake news detection while maintaining computational efficiency.