Tools
Tools
Ramesh, Naveen Kumar (2024) Comparison of Ensemble Techniques: Stacking vs. Voting Classifiers for Robust Fake News Detection on Social Media Using Deep Learning and NLP. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (1MB) | Preview |
Preview |
PDF (Configuration Manual)
Download (1MB) | Preview |
Abstract
This research investigates the effectiveness of ensemble learning mechanisms, stacking and voting classifiers, in detecting fake news disseminated through social media platforms via deep learning and natural language processing (NLP). With the increased spread of misinformation, building up strong detection systems able to adapt and operate high survivability towards complex text patterns has become a necessity. With a dataset of 23,481 fake news articles compared against 21,417 real news articles, the study undertakes sophisticated pre-processing techniques such as tokenization, stemming, and TF-IDF vectorization to prepare the data for classification. Basic learners include logistic regression, random forest, SVC and LSTM while the ensemble approaches will be evaluated on accuracy, precision, and computational efficiency. Results reveal that stacking classifiers surpasses voting classifiers with a logistic regression achieving the ultimate 94.07% as a meta-model. This analysis brings the promise of using ensemble techniques in combating misinformation and indicates some scope for the design of scalable and efficient systems for detecting fake news on social media platforms. Future work could include but not be limited to multimodal data, more advanced architectures such as transformer-based models, real-time applications, etc.
Actions (login required)
 |
View Item |