Spam Email Detection Using Deep Learning Techniques

Metadane

• Autorzy: I. AbdulNabi, Q. Yaseen
• Rok: 2021
• Źródło: Procedia Computer Science, Volume 184, Pages 853–858
• DOI/Link: 10.1016/j.procs.2021.03.107
• Status: to-read
• Pochodzenie: Wyekstrahowane z al-subaiey-web-ai-phishing-2024 ([24] - BERT transformer, Table 3 comparison)
• Tagi: to-read reference bert transformer spam-detection deep-learning fine-tuning

Notatki

Publikacja dodana automatycznie z bibliografii.

Kontekst cytowania w Al-Subaiey 2024:

• Referenced jako [24] w paper (Table 3 literature comparison)
• Method: Fine-tuned BERT transformer
• Dataset: Spam Base + Spam Filter Data (Kaggle) - 5,000 emails (3,000 spam, 2,000 ham)
• Results: Accuracy 98.67%, F1-score 98.66%
• Comparison: Al-Subaiey SVM+TF-IDF (99.1% acc, 99% F1) outperforms despite BERT complexity

Key Insight from Al-Subaiey:

• Traditional ML (SVM+TF-IDF) competitive with deep learning (BERT) for email classification
• Simpler models may be preferable: lower computational cost, better interpretability
• Dataset size: 5k vs Al-Subaiey 82k (scalability advantage for traditional ML)

Significance:

• Demonstrates state-of-art transformer application
• Benchmark for comparing traditional ML vs deep learning
• Smaller dataset typical of prior work → Al-Subaiey addresses generalizability gap

Dodaj PDF aby wygenerować pełne podsumowanie używając /summarize-paper abdulnabi-bert-spam-detection-2021