Fine-tuned RoBERTa-base for phishing email detection. Trained on 284k emails. Achieves 99.78% accuracy and 0.9963 F1 on a held-out test set of 21,773 emails.
Evaluation Results
Metric
Score
Accuracy
99.54%
F1 Score
0.9934
Training Data
Trained on 8 datasets totaling 284k emails including CEAS_08, Enron, Ling, Nazario, Nigerian_Fraud, SpamAssasin, and two Kaggle datasets.
Usage
from transformers import pipeline
classifier = pipeline("text-classification", model="eduardocastellon/roberta-phishing-email-detector")
result = classifier("Click here to claim your prize!")
print(result)
