부산대학교 도서관

Efficient grouping of Diagnosis Related Groups (DRGs) plays a crucial role in medical resource management and cost control, ultimately impacting the quality of patient care. To accurately assign patients to appropriate DRGs, it is essential to have precise diagnoses and surgical procedure information. In this study, we propose a novel methodology that leverages semantic analysis of clinical notes to enhance DRGs grouping prediction. We employ the powerful combination of Bidirectional Encoder Representations from Transformers (BERT) and Light Gradient Boosting Machine (LightGBM) algorithms. BERT enables us to capture the contextual semantic information inherent in the text data, while LightGBM provides a robust framework for predicting disease diagnoses and surgical procedures. Through extensive experiments and evaluations, we demonstrate the superior performance of our approach, surpassing traditional methods with an impressive F1-Score of over 91%. Our research not only contributes to natural language text modeling in the medical field but also provides valuable insights for the rational allocation and effective management of medical resources.