Development of an Ayurveda-Integrated Feature Engineering Framework for Disease Prediction

Abstract

A combination of the conventional Ayurvedic diagnostic knowledge and the recent computational intelligence should provide a direction to an improved way of improving the accuracy of diagnosing diseases and broadening the horizon of the entire healthcare provision. This paper introduces an Ayurveda-Based Feature Engineering (AFE) Framework in disease prediction with the assistance of machine-learning techniques. The systematically Ayurvedic diagnostic parameters of the Ayurvedic Prakriti constitution, Dosha imbalance, Agni condition, Nadi, and Astavidha Pariksha are systematically translated into structured machine-readable numerical features. To create a high-quality set of features that was consistent with the traditional medical reasoning and data science demands, a dataset gathered in the Ayurvedic hospitals and clinics was annotated with these parameters encoded. Several machine learning classifiers such as the random forest (RF), the support vector machine (SVM) as well as the navie bayes (NB) were trained and optimized using this improved dataset. Experiments indicate that using Ayurveda elements of diagnoses leads to a significant increase in predictive performance over traditional symptom-only models with significant improvements in accuracy, F1-score, and AUC measures. The presented AFE framework contributes to a powerful bridge between Ayurveda classical and contemporary predictive analytics and makes it possible to implement culturally-rooted and predictable disease-based forecasting systems. The contribution provides the foundation of the future study on integrative healthcare analytics and provides a scalable framework of building more sophisticated Ayurveda-informed clinical decision support systems.