Fake news stance detection using selective features and FakeNET

Area of Science:

• Computational Linguistics
• Artificial Intelligence
• Natural Language Processing

Background:

• The rapid spread of online information necessitates automated fake news detection systems.
• Effective fake news detection relies on robust feature engineering and dimensionality reduction.
• Existing methods face challenges in performance and computational complexity.

Purpose of the Study:

• To develop an automated system for timely fake news judgment.
• To evaluate feature dimensionality reduction techniques (Chi-square and PCA) for fake news detection.
• To assess the performance of a hybrid Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) model (FakeNET) using reduced feature sets.

Main Methods:

• Utilized a hybrid neural network architecture: FakeNET (CNN-LSTM).
• Employed Chi-square and Principal Component Analysis (PCA) for feature dimensionality reduction.
• Trained and evaluated the model on a multi-class dataset ('agree', 'disagree', 'discuss', 'unrelated') from the Fake News Challenges (FNC).

Main Results:

• Principal Component Analysis (PCA) achieved a higher accuracy of 0.978 compared to Chi-square and state-of-the-art methods.
• The proposed approach demonstrated gains of 0.04 in accuracy and 0.20 in F1 score.
• PCA and Chi-square provided contextual features with nonlinear characteristics for improved fake news identification.

Conclusions:

• Feature dimensionality reduction using PCA is effective for enhancing fake news detection performance.
• The FakeNET architecture, combined with PCA, offers a robust and computationally efficient solution for classifying news stance.
• The study highlights the importance of appropriate feature selection for building accurate automated fake news detection systems.