Utilizing Pretrained GPT-2 Model for Analyzing Drug Reviews


Creative Commons License

Özdemir S., Turhan K.

15. Tıp Bilişimi Kongresi, Trabzon, Türkiye, 30 - 31 Mayıs 2024, ss.306

  • Yayın Türü: Bildiri / Özet Bildiri
  • Basıldığı Şehir: Trabzon
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.306
  • Karadeniz Teknik Üniversitesi Adresli: Evet

Özet

 Accurately predicting drug side effects is critical to ensuring patient safety and optimizing health outcomes, while drug efficacy prediction plays a crucial role in directing clinicians and patients to treatments that provide maximum benefit with minimal side effects. This study used the pre-trained GPT-2 model by fine tuning to analyze drug reviews. Our aim is to estimate side effects in three categories (mild, severe and no side effects) and to classify effectiveness (ineffective, moderately effective and highly effective) into three categories. In conclusion, drug effectiveness and drug side effects prediction based on drug reviews represents a promising paradigm in precision medicine. By integrating advanced computational methods with clinical data, It can improve our understanding of drug responses and provide personalized health care by tailoring treatments to individual patient needs.