Smart Thyroid: A Deep Learning-Driven SVM Model for Thyroid Diagnosis
M. Anitha, R Keerthana, E Regunath, R Hariram · 2025
Effective therapy and patient care for thyroid conditions like hypothyroidism and hyperthyroidism depend on an accurate and timely diagnosis. Automated and accurate categorization algorithms are necessary since traditional diagnostic methods frequently face difficulties such data imbalance, overlapping symptoms, and subjective interpretation. An AI-powered real-time thyroid illness prediction system that makes use of cutting-edge machine learning techniques is presented in this paper. In order to rectify class imbalance and guarantee a more representative dataset for model training, the Synthetic Minority Over-Sampling Technique (SMOTE) is utilized. Thyroid disorders that can be reliably diagnosed and classified using K-Nearest Neighbors (KNN), Decision Tree, and Deep Support Vector Machine (Deep SVM) models include normal function, hypothyroidism, hyperthyroidism, and Graves' disease. These models are compared using key performance metrics such as F1-score, recall, accuracy, and precision. With an accuracy of 97.35%, experimental findings show that the Deep SVM model performs better than other classifiers, guaranteeing increased diagnostic reliability. The suggested system helps medical practitioners make data-driven clinical decisions by offering a reliable, automated, and scalable method of detecting thyroid disease.