AI-Driven Health Support Chatbot Using Retrieval-Augmented Generation (RAG)

Abstract

The increasing adoption of digital healthcare services has created a strong need for intelligent systems that can deliver dependable and interpretable preliminary medical support. This paper proposes an AI-based healthcare assistance chatbot that integrates machine learning–driven disease prediction with a Retrieval-Augmented Generation (RAG) framework to provide accurate, context-aware, and explainable health guidance. The proposed system allows users to describe their symptoms in natural language, which are analyzed using advanced text processing techniques to extract clinically relevant indicators.

A supervised learning–based classification model is employed to predict the most likely disease conditions from the extracted symptoms. To enhance the reliability and safety of the generated responses, a RAG architecture retrieves validated medical knowledge from a structured repository, including

disease profiles, common symptoms, severity levels, and recommended precautionary measures. This retrieved information is subsequently combined with a generative language model to produce coherent, evidence-supported, and user-friendly health explanations.

By grounding the generative output in verified medical data, the proposed approach significantly reduces hallucination risks and improves transparency in automated health assistance. The system not only supports preliminary symptom assessment but also delivers educational insights and promotes timely consultation with healthcare professionals. The experimental results demonstrate that combining machine learning–based disease prediction with retrieval-driven knowledge grounding substantially improves the accuracy, trustworthiness, and overall quality of digital healthcare support systems.