본문 바로가기
  • Home

SMART-MDR: Stacked Multimodal Architecture with Robust Text-Tabular Data for Medical Department Recommendation

  • Journal of The Korea Society of Computer and Information
  • Abbr : JKSCI
  • 2025, 30(12), pp.245~252
  • Publisher : The Korean Society Of Computer And Information
  • Research Area : Engineering > Computer Science
  • Received : October 31, 2025
  • Accepted : December 10, 2025
  • Published : December 31, 2025

Yang-Hoon Ham 1 Seong-Min Lee 1 Min-soo Kim 1 Chanhee Kwak 1

1강남대학교

Accredited

ABSTRACT

This study proposes a multimodal system that integrates subjective symptom text with structured data to recommend the most appropriate medical department. 2.89 million medical records provided by the National Health Insurance Service, was used to determine key diagnosis codes and medical department data. During data preprocessing, irrelevant departments were excluded, and data augmentation was performed using a large language model (LLM) to enhance data quality. The proposed model was designed with a dual-track architecture combining KM-BERT for text data processing and XGBoost for structured data analysis, with a CatBoost-based Stack-Ensemble model for final prediction. Experimental results showed that the proposed model achieved an accuracy of 0.748 and an F1-score of 0.732 compared to existing single models.

Citation status

* References for papers published after 2024 are currently being built.