Detection of Fake News Text Classification on COVID-19 Using Deep Learning Approaches.

Area of Science:

• Natural Language Processing
• Machine Learning
• Social Media Analytics

Background:

• Social media generates vast data, with Twitter being a key platform for public opinion.
• Sentiment analysis (SA) of this data, particularly for identifying fake news on critical topics like COVID-19, presents significant challenges.
• Accurate SA is crucial for understanding public discourse and combating misinformation.

Purpose of the Study:

• To propose a highly accurate sentiment analysis approach for fake news related to COVID-19 on Twitter.
• To evaluate and compare the performance of various machine learning and deep learning algorithms for this task.
• To develop an efficient prediction model for classifying the sentiment of COVID-19 fake news.

Main Methods:

• Data preprocessing including handling missing values, noise removal, tokenization, and stemming.
• Application of a semantic model with Term Frequency-Inverse Document Frequency (TF-IDF) weighting for data representation.
• Evaluation of eight machine learning algorithms (Naive Bayesian, Adaboost, KNN, Random Forest, Logistic Regression, Decision Tree, Neural Networks, SVM) and four deep learning models (CNN, LSTM, RNN, GRU).

Main Results:

• The developed prediction model demonstrated high accuracy in classifying COVID-19 fake news sentiments.
• The proposed approach outperformed other evaluated models in sentiment analysis tasks.
• Performance was rigorously assessed using metrics like confusion matrix, classification rate, and true positive rate.

Conclusions:

• The research successfully developed an efficient and highly accurate sentiment analysis model for COVID-19 fake news on Twitter.
• The findings provide valuable insights for researchers in selecting effective SA models for social media data.
• Recommendations and future research directions are offered to advance the field of fake news detection and sentiment analysis.