An Intelligent Medical Chatbot for Symptom Severity Classification and Personalized Recommendation Using Random Forest
Harshada Chavan∗, Darshan Mandavkar∗, Atharva Jagtap∗, Maitreya Kokane∗, Dr. Kishor Sakure†
∗Department of Computer Engineering, Terna Engineering College, University of Mumbai, Navi Mumbai, India
†Project Guide, Department of Computer Engineering,
Terna Engineering College, University of Mumbai, Navi Mumbai, India
Abstract—The rapid expansion of digital health technologies has created a pressing need for intelligent, accessible triage and diagnostic assistance systems. This paper presents a Medical Chatbot platform that leverages machine learning to interact with patients, classify symptom severity, and provide comprehensive medical recommendations. The system utilizes a conversational framework built on Streamlit with text-based symptom ex- traction, allowing seamless analysis of self-reported symptom intensities. By employing a Random Forest Classifier trained on synthetic medical data infused with Gaussian noise, the system accurately categorizes conditions into Normal, Moderate, or High severity levels. The platform further integrates a secure SQLite- backed authentication module with bcrypt password hashing and a rule-based recommendation engine for medication, diet, and appropriate precautions spanning Ayurvedic, Homeopathy, and Allopathy treatment methodologies. Experimental evaluation demonstrates balanced training and testing accuracies (averaging 75%–80%), indicating robust generalization without overfitting. The proposed system bridges the gap between initial symptom onset and formal medical consultation by providing continuous, data-driven medical triage. Results confirm that the integration of machine learning within interactive chatbots significantly enhances accessible preliminary healthcare.
Index Terms—Artificial Intelligence, Medical Chatbot, Ma- chine Learning, Random Forest Classifier, Symptom Severity Classification, Streamlit, Telemedicine, Triage Engine
