Comparative Study of Fake News Detection Using Sentiment-Integrated Logistic Regression, LSTM, and Hybrid Models
Zihao Nie · Applied and Computational Engineering · 2024
With the advent of the digital age, fake news spreads faster and faster on social media, causing a major adverse impact on social public opinion. Therefore, effective information detection technology is very important to mitigate the negative impact of false information and protect the health and stability of society. This research aims to improve fake news detection by incorporating sentiment analysis into traditional machine learning and deep learning models. The dataset used ISOT dataset contains more than 40,000 news articles and is used to compare the performance of logistic regression, Long Short-Term Memory (LSTM), and hybrid ensemble models. The study results show that the accuracy of the integrated model is the highest, reaching 99.24%, and the F1 value is 0.9922. The accuracy of logistic regression also reached 99.12%. Although sentiment analysis can add some value, it has a limited impact on model performance. This means that combining the traditional learning model with the deep learning model can enhance the fake news detection effect.