Shah, Samrat Sanjaykumar (2024) Email Spam Detection: Leveraging Fine-Tuned Transformer Models with Attention Mechanism. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (1MB) | Preview |
Preview |
PDF (Configuration Manual)
Download (2MB) | Preview |
Abstract
Due to ongoing threats to email security, it is becoming increasingly important to use advanced methods to consistently get rid of unwanted emails. To meet this need three advanced machine learning (ML) techniques DistilBERT, XLM-RoBERTa, and RoBERTa are tested to see how well they can find spam emails. Along with that pre-trained ML systems are tuned on the Enron-Spam dataset, which is a standard way to test how well spam identification works. Metrics like accuracy, precision, recall, and F1-score are used to test and analyze these improved systems in great depth to see how well they work. The research also investigates how focusing features built into these designs can make the models more accurate and clearer. The results show that the best method is the improved DistilBERT model, which is 96% accurate. The study shows that focusing mechanisms are important for making these models work better by helping with more accurate feature extraction and classification. Furthermore, this study adds to the progress in email security by showing how advanced ML can be used to find spam and how important narrowing methods are for making models work better. These findings are important for making spam filtering technologies better and more reliable. This will improve email security and the user experience in today's digital world.
Actions (login required)
![]() |
View Item |