Browsing by Author "Vega Rivera, Paulina Valeria"

Now showing 1 - 1 of 1

Comparative analysis of deep learning models versus machine learning algorithms for spam email detection
(Universidad Técnica Federico Santa María, 2025-07) Vega Rivera, Paulina Valeria; Ñanculef Alegría, Juan Ricardo; Departamento de Informática; Valle Vidal, Carlos Antonio
Spam email detection is a critical task in cybersecurity, complicated by constantly evolving spam tactics that are designed to trick the filtering systems. This study compares five machine learning models and nine deep learning models, evaluated using eight performance metrics on a combined dataset of 5,500 email. Statistical tests were applied to the top-performing models to asses significance. Results show that RoBERTa consistently achieves the highest F1 score among all deep learning models, while the fine-tuned GPT models, considered a special case due to being trained on significantly smaller datasets, still perform competitively. Among machine traditional learning models, SVM, NB and RF achieved the highest score, however, they still performed worse than the five Transformer-based models. Overall, the study's goal is to provide a comprehensive benchmark of traditional and modern approaches to spam detection under practical constraints.