// <![CDATA[SISTEM REKOMENDASI POLI KLINIK DI RSUD AJI BATARA AGUNG DEWA SAKTI BERBASIS RETRIEVAL-AUGMENTED GENERATION DAN LARGE LANGUAGE MODEL]]> 120160503 - Nur Fitrianti Fahrudin, S.Kom., M.T. Dosen Pembimbing 1 Muhammad Fajar Maulana Madani / 162021012 Penulis
Rumah Sakit Umum Daerah (RSUD) Aji Batara Agung Dewa Sakti menghadapi tantangan ketika pasien, khususnya pasien baru, kesulitan menentukan poli klinik yang tepat sesuai dengan keluhan medis. Kondisi ini menimbulkan risiko salah rujukan, keterlambatan diagnosis, serta peningkatan beban kerja staf informasi. Penelitian ini mengusulkan pembangunan sistem rekomendasi poli klinik berbasis Retrieval-Augmented Generation (RAG) dan Large Language Model (LLM) untuk membantu pasien memilih poli klinik yang sesuai dengan kondisi mereka. Data yang digunakan berasal dari e-rekam medis RSUD periode 2024, mencakup atribut demografis dan diagnosis pasien. Tahapan penelitian meliputi data preprocessing (data cleaning, normalisasi, ekspansi singkatan medis, stopword removal, serta feature augmentation menggunakan LLM), embedding menggunakan Sentence Transformer, penyimpanan dalam ChromaDB, dan penerapan similarity search dengan algoritma Approximate Nearest Neighbor (HNSW) berbasis cosine distance. Sistem kemudian mengintegrasikan hasil retrieval dengan LLM (LLaMA) untuk menghasilkan rekomendasi poli klinik. Evaluasi dilakukan melalui expert-based evaluation yang melibatkan tenaga medis sebagai validator. Hasil evaluasi menunjukkan bahwa sistem mampu memberikan rekomendasi yang akurat, relevan, dan kontekstual sesuai kondisi pasien. Implementasi prototipe ini diharapkan dapat meningkatkan efisiensi pelayanan rumah sakit, mengurangi beban staf informasi, serta membantu pasien memperoleh rujukan poli klinik yang lebih tepat. Selain itu, penelitian ini juga memberikan kontribusi terhadap pemanfaatan teknologi NLP berbasis RAG dalam sistem kesehatan daerah. Aji Batara Agung Dewa Sakti Regional General Hospital (RSUD) faces a recurring challenge where patients, particularly new ones, experience difficulties in selecting the appropriate outpatient clinic for their medical conditions. This often leads to misreferrals, delayed diagnoses, and increased workload for hospital staff. To address this issue, this study proposes the development of a clinic recommendation system based on Retrieval-Augmented Generation (RAG) and Large Language Models (LLM) to assist patients in determining the most relevant clinic. The dataset used consists of electronic medical records (EMR) from RSUD for the year 2024, covering demographic attributes and patient diagnoses.The research methodology involves several stages, including data preprocessing (data cleaning, normalization, expansion of medical abbreviations, stopword removal, and feature augmentation using LLM), embedding generation with Sentence Transformer, storage in ChromaDB, and similarity search using the Approximate Nearest Neighbor (HNSW) algorithm with cosine distance. The retrieved results are then integrated with an LLM (LLaMA) to generate clinic recommendations. Evaluation was conducted using expert-based evaluation, involving medical professionals as validators. The findings indicate that the proposed system is capable of producing accurate, relevant, and contextually appropriate recommendations according to patient conditions.The implementation of this prototype is expected to enhance hospital service efficiency, reduce the burden on hospital staff, and assist patients in obtaining more accurate clinic referrals. Furthermore, this research contributes to the application of RAG-based NLP technology in the healthcare sector, particularly in regional hospitals.