• 

  Result

Cancer_LLM is a research-oriented project for building high-quality text datasets focused on cancer outcomes, prognosis, and clinical research. It provides scripts to scrape, clean, and organize oncology articles into structured text that can be used for medical large language model (LLM) training and AI research. The core idea is to gather research articles on various cancer types from public sources (like PubMed Central), extract the critical sections of the text (title, abstract, body), remove noise, and save them in a unified format suitable for LLM training.

BioBERT Cancer Classification � Test Results

Model : dmis-lab/biobert-base-cased-v1.2 Classes : ['Blood_Cancer', 'Brain_Cancer', 'Liver_Cancer', 'Lung_Cancer', 'Mouth_Cancer', 'Ovarian_Cancer', 'Protaste_Cancer', 'Skin_Cancer']

             precision    recall  f1-score   support

Blood_Cancer 0.95 0.96 0.95 598 Brain_Cancer 0.93 0.95 0.94 617 Liver_Cancer 0.92 0.86 0.89 590 Lung_Cancer 0.91 0.93 0.92 631 Mouth_Cancer 0.98 0.98 0.98 587 Ovarian_Cancer 0.89 0.93 0.91 590 Protaste_Cancer 0.86 0.85 0.85 620 Skin_Cancer 0.92 0.90 0.91 567

   accuracy                           0.92      4800
  macro avg       0.92      0.92      0.92      4800

weighted avg 0.92 0.92 0.92 4800

Built With

• biobert
• deeplearning
• finetuning
• python