E-posta Kaynaklı Siber Riskleri Azaltmak için Yapay Zeka Tabanlı Sınıflandırma Modellerinin Geliştirilmesi ve Karşılaştırmalı Analizi

Abstract

Kurumsal elektronik posta (e-posta) sistemleri, işletmelerin iletişim ve iş süreçlerinin temel bileşenlerinden biri olmakla birlikte, spam ve kimlik avı (phishing) saldırıları gibi önemli siber güvenlik risklerini de beraberinde getirmektedir. Özellikle artan dijitalleşme ile birlikte işletmelerin maruz kaldığı e-posta tabanlı tehditlerin doğru ve hızlı şekilde tespit edilmesi kritik bir gereklilik haline gelmiştir. Bu araştırmada, spam, normal ve kimlik avı mesajlarının otomatik olarak sınıflandırılması yaklaşımına yönelik makine öğrenmesi ve transformer tabanlı derin öğrenme modellerinin karşılaştırmalı performans analizi gerçekleştirilmiştir. Araştırma kapsamında yaklaşık 515 bin İngilizce e-posta kaydından oluşan büyük ölçekli ve üç sınıflı bir veri seti kullanılmıştır. Çalışmada geleneksel makine öğrenmesi algoritmalarından Destek Vektör Makineleri (SVM), Lojistik Regresyon (LR), Naive Bayes (NB) ve Karar Ağacı (DT) modelleri ile transformer tabanlı BERT, RoBERTa ve ELECTRA modelleri değerlendirilmiştir. Kullanılan veri setine metin temizleme, stopword filtreleme, tokenleştirme ve vektörleştirme gibi veri ön işleme süreçleri uygulanmıştır. Modellerin performansları doğruluk (accuracy), kesinlik (precision), geri çağırma (recall), F1-Score, ROC-AUC ve confusion matrix ölçütleri kullanılarak analiz edilmiştir. Araştırma sonuçları, transformer tabanlı modellerin tüm performans metrikleri açısından geleneksel makine öğrenmesi modellerine kıyasla daha başarılı sonuçlar ürettiğini göstermektedir. En yüksek performans ELECTRA modeli tarafından elde edilmiş olup model %98,81 doğruluk ve 0,9837 F1-Score değerine ulaşmıştır. Diğer modeller ise %90-%98 aralığında değişen oranlarda performans sergilemiştir. Bulgular, transformer tabanlı modellerin özellikle phishing içeriklerinin tespitinde daha güçlü sınıf ayrım kapasitesi sunduğunu ortaya koymaktadır. Araştırmanın büyük ölçekli ve üç sınıflı veri yapısıyla ve farklı yapay zekâ mimarileri kullanılarak geliştirilen modelleriyle e-posta güvenliği ve yapay zekâ tabanlı siber risk yönetimi literatürüne katkı sağladığı değerlendirilmektedir.

Keywords

• E-posta sınıflandırması
• doğal dil işleme
• makine öğrenimi
• derin öğrenme
• yapay zeka.

Ethical Statement

Hazırlanan makalenin etik kurul izni ve/veya yasal/özel izin gerektirmeyen bir çalışma olduğunu beyan ederim.

Development and Comparative Analysis of AI-Based Classification Models for Mitigating Email-Related Cyber Risks

Abstract

Corporate email systems are a fundamental component of businesses' communication and business processes, but they also bring significant cybersecurity risks such as spam and phishing attacks. Especially as digitalization increases, the accurate and rapid detection of email-based threats faced by businesses has become a critical requirement. This research presents a comparative performance analysis of machine learning and transformer-based deep learning models for the automatic classification of spam, normal, and phishing messages. A large-scale, three-class dataset consisting of 479,466 English email records was used in the study. Traditional machine learning algorithms such as Support Vector Machines (SVM-LinearSVC), Logistic Regression (LR), Naive Bayes (NB), and Decision Tree (DT) models, as well as transformer-based BERT, RoBERTa, and ELECTRA models, were evaluated. The performance of the models was analyzed using accuracy, precision, recall, F1-Score, ROC-AUC, and confusion matrix metrics. In addition, statistical comparisons, phishing-specific security errors, cost-sensitive scenarios, and computational efficiency were evaluated to provide a more comprehensive assessment of model performance. Transformer-based models generally achieved stronger overall and class-specific classification performance, while conventional machine learning models offered substantially lower computational requirements. BERT achieved the highest overall performance, with 98.84% accuracy and a Macro F1-Score of 0.9827, while RoBERTa produced statistically comparable results. Security-oriented and computational analyses further showed that the model with the highest aggregate performance was not necessarily optimal for all deployment scenarios, highlighting the importance of considering misclassification consequences, computational requirements, email traffic volume, and available infrastructure in model selection. It is considered that the research contributes to the literature on email security and AI-based cyber risk management through models developed using a large-scale, three-class dataset, diverse AI architectures, and security-oriented evaluation criteria.

Keywords

• E-Mail classification
• natural language processing
• machine learning
• deep learning
• artificial intelligence.

Ethical Statement

I declare that the prepared article is a study that does not require ethics committee approval and/or legal/special permission.

References

1. [1] L. Jiang, ‘Variability of HR Practices and Their Impact on Organizational Commitment Across Different Sectors in South Africa’, J. Enterp. Bus. Intell., pp. 32–41, 2024, doi: 10.53759/5181/jebi202404004.
2. [2] M. C. Solano and J. C. Cruz, ‘Integrating Analytics in Enterprise Systems: A Systematic Literature Review of Impacts and Innovations’, Adm. Sci., vol. 14, no. 7, p. 138, 2024, doi: 10.3390/admsci14070138.
3. [3] Ö. Külcü, ‘Kurumsal Bilgi Kaynağı Olarak E-posta Yönetim Sistemleri’, Bilgi Dünya., vol. 13, no. 1, pp. 226–248, Apr. 2012, doi: 10.15612/BD.2012.176.
4. [4] E. E. E. Eryılmaz, D. Ö. Şahin, and E. Kılıç, ‘Türkçe İstenmeyen E-postaların Farklı Öznitelik Seçim Yöntemleri Kullanılarak Makine Öğrenmesi Algoritmaları ile Tespit Edilmesi’, Türkiye Bilişim Vakfı Bilgi. Bilim. Ve Mühendisliği Derg., vol. 13, no. 2, pp. 57–77, Dec. 2020.
5. [5] D. E. O’Leary, ‘KPMG Knowledge Management and the Next Phase: Using Enterprise Social Media’, J. Emerg. Technol. Account., vol. 13, no. 2, pp. 215–230, 2016, doi: 10.2308/jeta-51600.
6. [6] H. Karamollaoglu, İ. A. Dogru, and M. Dorterler, ‘Detection of Spam E-mails with Machine Learning Methods’, in 2018 Innovations in Intelligent Systems and Applications Conference (ASYU), Oct. 2018, pp. 1–5. doi: 10.1109/ASYU.2018.8554014.
7. [7] M. A. Koşan and R. Benzer, ‘Siber Güvenlik Alanında Derin Öğrenme Yöntemlerinin Kullanımı’, Jan. 2019, doi: 10.6084/m9.figshare.7581983.v1.
8. [8] E. Dönmez and N. K. Torun, ‘Siber cehennem: sağlık sektörü üzerine bir değerlendirme’, in Sosyal Bilimler Araştırmaları: Çıkarımlar, Bileşenler, Çözümlemeler, vol. 1, 1 vols, Klaipeda: SRA Academic Publishing, 2022, pp. 1–14.

9. [9] E. E. Eryılmaz, D. Özkan Şahin, and E. Kılıç, ‘Machine Learning Based Spam E-mail Detection System for Turkish’, in 2020 5th International Conference on Computer Science and Engineering (UBMK), 2020, pp. 7–12. doi: 10.1109/UBMK50275.2020.9219487.
10. [10] K. Çalış, O. Gazdağı, and O. Yıldız, ‘Reklam İçerikli Epostaların Metin Madenciliği Yöntemleri İle Otomatik Tespiti’, Bilişim Teknol. Derg., vol. 6, no. 1, pp. 1–7, May 2013.
11. [11] C. Yucel, A. Lockett, I. Chalkias, D. Mallis, and V. Katos, ‘MAIT: Malware Analysis and Intelligence Tool’, Inf. Secur. Int. J., vol. 50, pp. 49–65, 2021, doi: https://doi.org/10.11610/isij.5024.
12. [12] F. J. A. Serna, ‘The Legal Regulation of Spam: An International Comparative Study’, J. Innov. Digit. Mark., vol. 3, no. 1, pp. 3–13, 2022, doi: 10.51300/jidm-2022-44.
13. [13] C. Özdemi̇r, M. Ataş, and A. B. Özer, ‘Classification of Turkish spam e-mails with artificial immune system’, in 2013 21st Signal Processing and Communications Applications Conference (SIU), 2013, pp. 1–4. doi: 10.1109/SIU.2013.6531457.
14. [14] D. Triggs et al., ‘A Distributed Registry: Smartphone-Connected Database Driven by Patient-Reported Outcomes’, J. Vasc. Surg., vol. 72, no. 1, p. e259, 2020, doi: 10.1016/j.jvs.2020.04.428.
15. [15] N. H. A. B. Ghozali, N. S. H. NOZARI, and N. F. F. B. Zamzuri, ‘Types and Methods of Managing SPAM Messages: A Review’, 2023, doi: 10.36227/techrxiv.170327076.62289830/v1.
16. [16] H. Abroshan, J. Devos, G. Poels, and E. Laermans, ‘COVID-19 and Phishing: Effects of Human Emotions, Behavior, and Demographics on the Success of Phishing Attempts During the Pandemic’, Ieee Access, vol. 9, pp. 121916–121929, 2021, doi: 10.1109/access.2021.3109091.
17. [17] F. Carroll, J. A. Adejobi, and R. Montasari, ‘How Good Are We at Detecting a Phishing Attack? Investigating the Evolving Phishing Attack Email and Why It Continues to Successfully Deceive Society’, Sn Comput. Sci., vol. 3, no. 2, 2022, doi: 10.1007/s42979-022-01069-1.
18. [18] D. Jampen, G. Gür, T. Sutter, and B. Tellenbach, ‘Don’t Click: Towards an Effective Anti-Phishing Training. A Comparative Literature Review’, Hum.-Centric Comput. Inf. Sci., vol. 10, no. 1, 2020, doi: 10.1186/s13673-020-00237-7.
19. [19] D. S. Oliveira et al., ‘Empirical Analysis of Weapons of Influence, Life Domains, and Demographic-Targeting in Modern Spam: An Age-Comparative Perspective’, Crime Sci., vol. 8, no. 1, 2019, doi: 10.1186/s40163-019-0098-8.
20. [20] Y. Tian, X. Dai, Z. Li, H. Guo, and X. Mao, ‘Improving the Accuracy of Cybersecurity Spam Email Detection Using Ensemble Techniques: A Stacking Approach Machine Learning for Spam Email Detection’, Plos One, vol. 20, no. 9, p. e0331574, 2025, doi: 10.1371/journal.pone.0331574.
21. [21] B. B. Gupta, N. A. G. Arachchilage, and K. E. Psannis, ‘Defending Against Phishing Attacks: Taxonomy of Methods, Current Issues and Future Directions’, Telecommun. Syst., vol. 67, no. 2, pp. 247–267, 2018, doi: 10.1007/s11235-017-0334-z.
22. [22] S. Das, C. Nippert‐Eng, and L. J. Camp, ‘Evaluating User Susceptibility to Phishing Attacks’, Inf. Comput. Secur., vol. 30, no. 1, pp. 1–18, 2022, doi: 10.1108/ics-12-2020-0204.
23. [23] H. Fujima, K. Takeuchi, and T. Kumamoto, ‘Semantic Analysis of Phishing Emails Leading to Ransomware With ChatGPT’, 2023, doi: 10.21203/rs.3.rs-3670198/v1.
24. [24] K. F. Tschakert and S. Ngamsuriyaroj, ‘Effectiveness of and User Preferences for Security Awareness Training Methodologies’, Heliyon, vol. 5, no. 6, p. e02010, 2019, doi: 10.1016/j.heliyon.2019.e02010.
25. [25] L. Gallo, A. Maiello, A. Botta, and G. Ventre, ‘2 Years in the Anti-Phishing Group of a Large Company’, Comput. Secur., vol. 105, p. 102259, 2021, doi: 10.1016/j.cose.2021.102259.
26. [26] U. P. Sajad, ‘A Robust and Explainable Transformer-Based Framework for Phishing Email Detection’, 2025, doi: 10.48550/arxiv.2511.12085.
27. [27] M.-L. E. Alorvor and S. Dadkhah, ‘Real-Time Phishing Detection for Brand Protection Using Temporal Convolutional Network-Driven URL Sequence Modeling’, Electronics, vol. 14, no. 18, p. 3746, 2025, doi: 10.3390/electronics14183746.
28. [28] C. S. Wang, F. Hsu, S. J. Chen, Y. L. Hwang, and M. Wu, ‘Detecting Phishing Sites Using URLs Collected From Emails’, Appl. Mech. Mater., vol. 479–480, pp. 916–922, 2014, doi: 10.4028/www.scientific.net/amm.479-480.916.
29. [29] A. Karim, S. Azam, B. Shanmugam, K. Kannoorpatti, and M. Alazab, ‘A Comprehensive Survey for Intelligent Spam Email Detection’, Ieee Access, vol. 7, pp. 168261–168295, 2019, doi: 10.1109/access.2019.2954791.
30. [30] R. Tóth, T. Bisztray, and N. Gruschka, ‘The Phish, the Spam, and the Valid: Generating Feature-Rich Emails for Benchmarking LLMs’, 2026, doi: 10.48550/arxiv.2511.21448.
31. [31] B. Chan, S. Schweter, and T. Möller, ‘German’s Next Language Model’, Proc. 28th Int. Conf. Comput. Linguist. COLING 2020, pp. 6788–6796, 2020, doi: 10.18653/v1/2020.coling-main.598.
32. [32] C. Wang, Q. Li, T. Ren, X. Wang, and G. Guo, ‘High Efficiency Spam Filtering: A Manifold Learning-Based Approach’, Math. Probl. Eng., vol. 2021, p. 7, 2021, doi: 10.1155/2021/2993877.
33. [33] Y. Zhu, ‘Naive Bayesian Spam Filtering’, Highlights Sci. Eng. Technol., vol. 38, pp. 64–69, 2023, doi: 10.54097/hset.v38i.5734.
34. [34] I. Moutafis, A. Andreatos, and P. Stefaneas, ‘Spam Email Detection Using Machine Learning Techniques’, Eur. Conf. Cyber Warf. Secur., vol. 22, no. 1, pp. 303–310, 2023, doi: 10.34190/eccws.22.1.1208.
35. [35] S. Jamal, H. Wimmer, and I. H. Sarker, ‘An Improved Transformer‐based Model for Detecting Phishing, Spam and Ham Emails: A Large Language Model Approach’, Secur. Priv., vol. 7, no. 5, 2024, doi: 10.1002/spy2.402.
36. [36] A. Yıldız and M. Demirci, ‘Corporate E-Mail Classification System’, in 2017 International Conference on Computer Science and Engineering (UBMK), 2017, pp. 559–564. doi: 10.1109/UBMK.2017.8093462.
37. [37] F. Agusman, ‘Comparative study of DistilBERT and ELECTRA-Small Models in Spam Email Classification’, J. Inform., vol. 12, no. 2, pp. 113–121, 2025.
38. [38] İ. E. E. Parlak, ‘Kurumsal Şikayet Yönetimi için Makine Öğrenmesi Tabanlı Otomatik E-posta Sınıflandırma Sistemi’, J. Intell. Syst. Theory Appl., vol. 9, no. 2026, p. 1806772, Mar. 2026, doi: 10.38016/jista.1806772.
39. [39] M. Yağanoğlu and E. Irmak, ‘Separation of Incoming E-Mails Through Artificial Intelligence Techniques’, Avrupa Bilim Ve Teknol. Derg., no. 21, pp. 690–696, Jan. 2021, doi: 10.31590/ejosat.841299.
40. [40] A. Yücel and M. K. Köylü, ‘Spam İçerikli E-Postaların Tespiti için Bir Metin Madenciliği Uygulaması: Terimlerin Gama İlişki Katsayisina Dayalı Polarizasyonu’, Int. J. Manag. Inf. Syst. Comput. Sci., vol. 2, no. 2, pp. 95–104, 2018.
41. [41] M. Aydoğan and A. Karcı, ‘Spam Mail Detection Using Naive Bayes Method with Apache Spark’, in 2018 International Conference on Artificial Intelligence and Data Processing (IDAP), IEEE, 2018, pp. 1–6. doi: 10.1109/IDAP.2018.8620737.
42. [42] E. Şahin, M. Aydos, and F. Orhan, ‘Spam/Ham E-Mail Classification Using Machine Learning Methods Based on Bag of Words Technique’, in 2018 26th Signal Processing and Communications Applications Conference (SIU), May 2018, pp. 1–4. doi: 10.1109/SIU.2018.8404347.
43. [43] Z. A. Güven, ‘Türkçe E-postalarda Spam Tespiti için Makine Öğrenme Yöntemlerinin ve Dil Modellerinin Analizi’, Avrupa Bilim Ve Teknol. Derg., no. 47, pp. 1–6, Jan. 2023, doi: 10.31590/ejosat.1234079.
44. [44] J. V. Chandra, N. Challa, and S. K. Pasupuleti, ‘A practical approach to E-mail spam filters to protect data from advanced persistent threat’, 2016 Int. Conf. Circuit Power Comput. Technol. ICCPCT, pp. 1–5, 2016, doi: 10.1109/ICCPCT.2016.7530239.
45. [45] T. B. Shahi and A. Yadav, ‘Mobile SMS Spam Filtering for Nepali Text Using Naïve Bayesian and Support Vector Machine’, Int. J. Intell. Sci., vol. 4, no. 01, pp. 24–28, 2014, doi: 10.4236/ijis.2014.41004.
46. [46] M. Bassiouni, M. Ali, and E. El-Dahshan, ‘Ham and Spam E-Mails Classification Using Machine Learning Techniques’, J. Appl. Secur. Res., vol. 13, no. 3, pp. 315–331, 2018, doi: 10.1080/19361610.2018.1463136.
47. [47] S. A. Saab, N. Mitri, and M. Awad, ‘Ham or spam? A comparative study for some content-based classification algorithms for email filtering’, in MELECON 2014-2014 17th IEEE Mediterranean Electrotechnical Conference, IEEE, 2014, pp. 339–343. doi: 10.1109/MELCON.2014.6820574.
48. [48] V. Altıntaş, ‘E-Posta Dolandırıcılığının Tespi̇ti̇ içi̇n Hi̇bri̇t Nai̇ve Bayes ve Deri̇n Öğrenme Yaklaşımı’, Int. J. 3D Print. Technol. Digit. Ind., vol. 9, no. 3, pp. 476–487, Dec. 2025, doi: 10.46519/ij3dptdi.1725050.
49. [49] V. S. Tida and S. Hsu, ‘Universal spam detection using transfer learning of BERT model’, ArXiv Prepr. ArXiv220203480, 2022, doi: 10.48550/arXiv.2202.03480.
50. [50] B. Alshawi, A. Munshi, M. Alotaibi, R. Alturki, and N. Allheeib, ‘Classification of SPAM Mail Utilizing Machine Learning and Deep Learning Techniques’, Int. J. Inf. Technol. Secur., vol. 16, no. 2, pp. 71–82, 2024, doi: 10.59035/FPKO7430.
51. [51] A. Utku, ‘YouTube Yorumlarından Spam Tespitine Yönelik Makine Öğrenmesi ve Derin Öğrenme Yöntemlerinin Karşılaştırmalı Bir Analizi’, Bingöl Üniversitesi Tek. Bilim. Derg., vol. 6, no. 1, pp. 30–50, Jun. 2025, doi: 10.5281/zenodo.15719417.
52. [52] Y. Liao, D. Li, and N. Zhang, ‘Comparison of Interpolation Models for Estimating Heavy Metals in Soils Under Various Spatial Characteristics and Sampling Methods’, Trans. Gis, vol. 22, no. 2, pp. 409–434, 2018, doi: 10.1111/tgis.12319.
53. [53] Y. Okano, Y. Kase, and H. Okano, ‘A Set-Theoretic Definition of Cell Types With an Algebraic Structure on Gene Regulatory Networks and Application in Annotation of RNA-seq Data’, Stem Cell Rep., vol. 18, no. 1, pp. 113–130, 2023, doi: 10.1016/j.stemcr.2022.10.015.
54. [54] K. Namee, J. Polpinij, and B. Luaphol, ‘A Hybrid Approach for Aspect-Based Sentiment Analysis: A Case Study of Hotel Reviews’, Curr. Appl. Sci. Technol., vol. 23, no. 2, 2022, doi: 10.55003/cast.2022.02.23.008.
55. [55] Ö. Köksal and E. H. Yılmaz, ‘Improving Automated Turkish Text Classification With Learning‐based Algorithms’, Concurr. Comput. Pract. Exp., vol. 34, no. 11, 2022, doi: 10.1002/cpe.6874.
56. [56] M. E. Purbaya, D. P. Rakhmadani, M. P. Arum, and L. Z. Nasifah, ‘Implementation of N-Gram Methodology to Analyze Sentiment Reviews for Indonesian Chips Purchases in Shopee E-Marketplace’, J. Resti Rekayasa Sist. Dan Teknol. Inf., vol. 7, no. 3, pp. 609–617, 2023, doi: 10.29207/resti.v7i3.4726.
57. [57] N. N. Moon et al., ‘Natural Language Processing Based Advanced Method of Unnecessary Video Detection’, Int. J. Electr. Comput. Eng. Ijece, vol. 11, no. 6, pp. 5411–5419, 2021, doi: 10.11591/ijece.v11i6.pp5411-5419.
58. [58] Y. Song, X. Liu, and Z. Zhou, ‘A Comprehensive Review of Text Classification Algorithms’, J. Electron. Inf. Sci., vol. 9, no. 2, 2024, doi: 10.23977/jeis.2024.090205.
59. [59] N. S. Rixe et al., ‘The Development of a Novel Natural Language Processing Tool to Identify Pediatric Chest Radiograph Reports With Pneumonia’, Front. Digit. Health, vol. 5, 2023, doi: 10.3389/fdgth.2023.1104604.
60. [60] M. A. Raza, ‘Cyber Security and Data Privacy in the Era of E-Governance’, Soc. Sci. J. Adv. Res., vol. 4, no. 1, pp. 5–9, 2024.
61. [61] M. Veziroğlu, E. Veziroğlu, and İ. Ö. Bucak, ‘Performance Comparison Between Naive Bayes and Machine Learning Algorithms for News Classification’, 2024, doi: 10.5772/intechopen.1002778.
62. [62] O. O. A. Arzube, J. A. Garcí­a-Dí­az, J. Medina-Moreira, H. Luna-Aveiga, and R. Valencia‐García, ‘Comparing Deep-Learning Architectures and Traditional Machine-Learning Approaches for Satire Identification in Spanish Tweets’, Mathematics, vol. 8, no. 11, p. 2075, 2020, doi: 10.3390/math8112075.
63. [63] B. I. Kzar and H. H. Safi, ‘Systematic Review of Sentiment Analysis and Predict Sarcastic’, J. Al-Qadisiyah Comput. Sci. Math., vol. 15, no. 2, 2023, doi: 10.29304/jqcm.2023.15.2.1241.
64. [64] F. Rustam, I. Ashraf, A. Mehmood, S. Ullah, and G. S. Choi, ‘Tweets Classification on the Base of Sentiments for US Airline Companies’, Entropy, vol. 21, no. 11, p. 1078, 2019, doi: 10.3390/e21111078.
65. [65] T. Sabri, S. Bahassine, O. E. Beggar, and M. Kissi, ‘An Improved Arabic Text Classification Method Using Word Embedding’, Int. J. Electr. Comput. Eng. Ijece, vol. 14, no. 1, pp. 721–731, 2024, doi: 10.11591/ijece.v14i1.pp721-731.
66. [66] J. Atwan, M. Wedyan, Q. Bsoul, A. Hammadeen, and R. Alturki, ‘The Use of Stemming in the Arabic Text and Its Impact on the Accuracy of Classification’, Sci. Program., vol. 2021, p. 1367210, 2021, doi: 10.1155/2021/1367210.
67. [67] S. Saifullah, R. Dreżewski, F. A. Dwiyanto, A. S. Aribowo, Y. Fauziah, and N. H. Cahyana, ‘Automated Text Annotation Using a Semi-Supervised Approach With Meta Vectorizer and Machine Learning Algorithms for Hate Speech Detection’, Appl. Sci., vol. 14, no. 3, p. 1078, 2024, doi: 10.3390/app14031078.
68. [68] A. Conneau et al., ‘Unsupervised Cross-Lingual Representation Learning at Scale’, pp. 8440–8451, 2020, doi: 10.18653/v1/2020.acl-main.747.
69. [69] J. Koutsikakis, I. Chalkidis, P. Malakasiotis, and I. Androutsopoulos, ‘GREEK-BERT: The Greeks Visiting Sesame Street’, presented at the 1th Hellenic Conference on Artificial Intelligence (SETN 2020), 2020, pp. 110–117. doi: 10.1145/3411408.3411440.
70. [70] C. Lothritz, K. Allix, L. Veiber, T. F. Bissyandé, and J. Klein, ‘Evaluating Pretrained Transformer-Based Models on the Task of Fine-Grained Named Entity Recognition’, presented at the Proceedings of the 28th International Conference on Computational Linguistics, 2020, pp. 3750–3760. doi: 10.18653/v1/2020.coling-main.334.
71. [71] Y. Guo, Y. Ge, Y. Yang, M. A. Al-Garadi, and A. Sarker, ‘Comparison of Pretraining Models and Strategies for Health-Related Social Media Text Classification’, Healthcare, vol. 10, no. 8, p. 1478, 2022, doi: 10.3390/healthcare10081478.
72. [72] H. Tian et al., ‘The Prediction of Clinical Mastitis in Dairy Cows Based on Milk Yield, Rumination Time, and Milk Electrical Conductivity Using Machine Learning Algorithms’, Animals, vol. 14, no. 3, p. 427, 2024, doi: 10.3390/ani14030427.
73. [73] S. Zhang, X. Zhang, W. Zhang, and A. Søgaard, ‘Sociolectal Analysis of Pretrained Language Models’, presented at the In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, 2021, pp. 4581–4588. doi: 10.18653/v1/2021.emnlp-main.375.
74. [74] K. Clark, T. Luong, Q. V. Le, and C. Manning, ‘ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators’, in ICLR, 2020. [Online]. Available: https://openreview.net/pdf?id=r1xMH1BtvB
75. [75] J. B. Gelindon, R. M. A. Velasco, and D. D. Gante, ‘Credit Risk Evaluation in Banking and Lending Sectors Using Neural Network Model’, J. Corp. Finance Manag. Bank. Syst., no. 23, pp. 12–35, 2022, doi: 10.55529/jcfmbs23.12.35.
76. [76] A. Alhuzali, A. Alloqmani, M. Aljabri, and F. Alharbi, ‘In-Depth Analysis of Phishing Email Detection: Evaluating the Performance of Machine Learning and Deep Learning Models Across Multiple Datasets’, Appl. Sci., vol. 15, no. 6, p. 3396, 2025, doi: 10.3390/app15063396.
77. [77] R. Meléndez, M. Ptaszynski, and F. Masui, ‘Comparative Investigation of Traditional Machine-Learning Models and Transformer Models for Phishing Email Detection’, Electronics, vol. 13, no. 24, p. 4877, 2024, doi: 10.3390/electronics13244877.