Scholarlink

<p>Nowadays, a major challenge is the increasing distortion of information on social media platforms like WhatsApp. Ensuring the authenticity of user-generated content has become critical to prevent the spread of misinformation. In this paper, the authors propose a hybrid model that integrates machine learning (ML), deep learning (DL), and knowledge graph techniques to classify WhatsApp messages as real or fake. The approach begins with natural language processing (NLP) based preprocessing, dataset balancing, and Term Frequency–Inverse Document Frequency (TF-IDF) feature extraction. For classification, multiple ML algorithms such as support vector machine (SVM), logistic regression, decision tree, and XGBoost are evaluated, along with DL models including deep neural networks (DNN), long short-term memory (LSTM), gated recurrent unit (GRU), and fine-tuned Bidirectional Encoder Representations from Transformers (BERT). To further enhance classification performance, a knowledge graph built from high-frequency phrases is used in conjunction with a graph-based contextual scoring system. The suggested hybrid approach outperforms solo ML or DL models in terms of accuracy, precision, recall, and F1-score, according to experimental results on a real-world WhatsApp dataset.</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

Detecting Fake WhatsApp Messages with Knowledge Graphs using Machine Learning and Deep Learning

  • 期刊论文

  • 发表日期：

    2026年2月3日

查看全文
  • Sidhartha Sankar Pradhan,
  • Neetu Faujdar,
  • Subhendu Sekhar Sahoo,
  • Amit Das

摘要

Nowadays, a major challenge is the increasing distortion of information on social media platforms like WhatsApp. Ensuring the authenticity of user-generated content has become critical to prevent the spread of misinformation. In this paper, the authors propose a hybrid model that integrates machine learning (ML), deep learning (DL), and knowledge graph techniques to classify WhatsApp messages as real or fake. The approach begins with natural language processing (NLP) based preprocessing, dataset balancing, and Term Frequency–Inverse Document Frequency (TF-IDF) feature extraction. For classification, multiple ML algorithms such as support vector machine (SVM), logistic regression, decision tree, and XGBoost are evaluated, along with DL models including deep neural networks (DNN), long short-term memory (LSTM), gated recurrent unit (GRU), and fine-tuned Bidirectional Encoder Representations from Transformers (BERT). To further enhance classification performance, a knowledge graph built from high-frequency phrases is used in conjunction with a graph-based contextual scoring system. The suggested hybrid approach outperforms solo ML or DL models in terms of accuracy, precision, recall, and F1-score, according to experimental results on a real-world WhatsApp dataset.