AyurSage: An AI-Assisted Ayurvedic Clinical Decision Support System Using Canonical Disease-Dosha-Symptom Patterns

Jagveer Singh Bedi1, Aditi Verma1, Diksha Sharma1, Yashi Upadhyay1, Ayushi Sharma2*, Diwakar Shrivastava2, Anshika Singh2

1Department of Computer Science and Engineering, Hindustan College of Science and Technology (AKTU-064), Agra–Delhi Highway (NH-2), Churmura, Farah, Mathura, 281122, Uttar Pradesh, India.

2Faculty Mentor, Hindustan College of Science and Technology (AKTU-064), Agra–Delhi Highway (NH-2), Churmura, Farah, Mathura, 281122, Uttar Pradesh, India

 *Corresponding Author: ayushisharma.hcst@sgei.org

Abstract - Ayurveda emphasizes personalized diagnosis through the balance of three doshas: Vata, Pitta, and Kapha. While Artificial Intelligence (AI) has shown promise in healthcare, most AI-driven Ayurvedic systems rely on questionnaire-based or limited datasets without grounding in classical knowledge. This paper presents AyurSage, an AI-assisted Clinical Decision Support System (CDSS) that integrates canonical Ayurvedic knowledge with machine learning. A two-level data model is introduced, consisting of structured disease-dosha-symptom-treatment patterns extracted from classical texts and synthetic patient instances generated through controlled variation of contextual features. A five-phase pipeline transforms textual data into a structured dataset with 16 features and ~3000 records. A Random Forest classifier achieves perfect accuracy for dosha prediction, while a tuned ensemble model achieves ~95-97% accuracy for disease prediction. A knowledge-based recommendation system ensures treatment traceability to canonical sources. The system is deployed as a full-stack web application supporting real-time inference. AyurSage demonstrates a scalable, interpretable, and epistemologically grounded approach to AI in traditional medicine.

Key Words: Ayurveda, Clinical Decision Support System, Machine Learning, Explainable AI, Dosha Prediction, Text Mining