TriDoshaX: AI Based Panchakarma Treatment and Diet Recommendation System

Abstract: Ayurvedic medicine offers a prevention-oriented, constitution-based approach to personalized healthcare through the Tridosha framework—Vata, Pitta, and Kapha—collectively describing an individual’s Prakriti. Despite its clinical relevance, widespread adoption is hindered by the scarcity of trained practitioners, the inherent subjectivity of manual assessments, and the absence of scalable digital tools. This paper presents TriDoshaX, a web-based Clinical Decision Support System (CDSS) that implements Ayurvedic Prakriti classification through a hybrid architecture combining supervised machine learning with rule-based Ayurvedic scoring. A structured questionnaire capturing 26 physiological, psychological, and lifestyle attributes is used to collect user data, which is subsequently processed through an ensemble Random Forest classifier trained on approximately 5,000 records. The Random Forest model achieves an accuracy of 0.826, precision of 0.851, recall of 0.826, and F1-score of 0.812, outperforming Decision Tree, Support Vector Machine, and Logistic Regression baselines. The predicted Dosha type drives a personalized recommendation engine that suggests Panchakarma therapies, dietary plans, and lifestyle guidelines consistent with classical Ayurvedic principles. Additionally, a Retrieval-Augmented Generation (RAG) pipeline powers a conversational AI assistant capable of responding to Ayurveda-specific queries. TriDoshaX bridges traditional Ayurvedic principles with modern AI-assisted healthcare delivery, providing an accessible platform for preventive health management.

Keywords: Ayurveda; Tridosha; Prakriti Classification; Panchakarma; Random Forest; Clinical Decision Support System; Retrieval-Augmented Generation; Personalized Healthcare.