BERT Model Achieves 97% Accuracy in Classifying Mpox Research
A study on arXiv (2607.26700) applies multilabel classification to 14,590 Mpox research articles, categorizing them into topics like outbreaks, vaccination, and epidemiology. Among AI models tested, BERT achieved 97.05% accuracy, 97.67% micro F1, and 96.46% macro F1. SHAP analysis was used to interpret model decisions. The WHO reports increasing Mpox cases in some regions, underscoring the need for efficient research organization.
Key facts
- Study uses multilabel classification on 14,590 Mpox research articles
- BERT model achieved 97.05% accuracy
- Micro F1 score: 97.67%, macro F1 score: 96.46%
- SHAP used for model interpretability
- WHO reports increasing Mpox cases in some regions
- Research topics include outbreaks, vaccination, epidemiology
- Study aims to automate classification for efficient analysis
- Published on arXiv with ID 2607.26700
Entities
Institutions
- World Health Organization
- arXiv