A Hybrid Machine Learning and Deep Learning Framework for Fake News Detection Using Sentiment Analysis

The rapid growth of social media platforms has significantly increased the spread of fake news, creating serious challenges for information credibility and public trust. Automated fake news detection has therefore become an important research area in data science and management studies. This paper presents a hybrid machine learning and deep learning framework for fake news detection using sentiment analysis and textual features. Traditional machine learning classifiers such as Naïve Bayes and Logistic Regression are combined with deep learning models including Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and Bidirectional LSTM (Bi-LSTM). Sentiment polarity is extracted using the VADER sentiment lexicon, while TF-IDF and word embedding techniques are employed for feature representation. Experimental evaluation is conducted on a Twitter-based fake news dataset obtained from Kaggle. The results show that Logistic Regression achieved an accuracy of 91.16%, while the proposed CNN–BiLSTM hybrid model achieved a testing accuracy of 89.94%, demonstrating improved robustness in identifying fake news. The findings indicate that integrating sentiment features with deep learning architectures provides an effective solution for combating misinformation on social media platforms.