Scholarlink

<p>Warts, a prevalent skin disease, usually stem from an infection with the Human Papillomavirus (HPV). Accurate treatment prediction remains challenging due to the imbalance in medical data distribution. This study introduces a transformer learning system by utilizing axiomatic fuzzy set (AFS) theory and a GPT-2-based tabular data generation technique called realistic relational and tabular transformer (REaLTabFormer), designed to assist physicians in choosing appropriate wart treatment methods. The REaLTabFormer is utilized to handle the imbalanced data by generating synthetic samples for enhancing the predictive accuracy of the model. Meanwhile, an AFS decision tree is constructed to provide descriptive information about the patients, allowing prediction of their response towards the treatment. The AFS decision tree provides valuable insights into patient characteristics, aiding in treatment selection. Experimental results demonstrate that the proposed approach achieves high average prediction accuracy, with 98.88% for cryotherapy and 95.66% for immunotherapy, outperforming traditional machine learning baselines. The proposed prediction system can aid physicians in selecting the most suitable treatment for wart patients, thereby improving the efficiency of diagnosis and treatment.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A transformer-based learning system for wart disease treatment based on axiomatic fuzzy set theory

  • 期刊论文

  • 发表日期：

    2025年7月28日

查看全文
  • Wenjuan Jia,
  • Jianshuai Zhang,
  • Hongyue Guo,
  • Lidong Wang,
  • Xiaodong Liu

摘要

Warts, a prevalent skin disease, usually stem from an infection with the Human Papillomavirus (HPV). Accurate treatment prediction remains challenging due to the imbalance in medical data distribution. This study introduces a transformer learning system by utilizing axiomatic fuzzy set (AFS) theory and a GPT-2-based tabular data generation technique called realistic relational and tabular transformer (REaLTabFormer), designed to assist physicians in choosing appropriate wart treatment methods. The REaLTabFormer is utilized to handle the imbalanced data by generating synthetic samples for enhancing the predictive accuracy of the model. Meanwhile, an AFS decision tree is constructed to provide descriptive information about the patients, allowing prediction of their response towards the treatment. The AFS decision tree provides valuable insights into patient characteristics, aiding in treatment selection. Experimental results demonstrate that the proposed approach achieves high average prediction accuracy, with 98.88% for cryotherapy and 95.66% for immunotherapy, outperforming traditional machine learning baselines. The proposed prediction system can aid physicians in selecting the most suitable treatment for wart patients, thereby improving the efficiency of diagnosis and treatment.