Thyroid Disease Multi-Class Classification
Gaurav A. Singh, Vaishnavi V. Mane, Rishabh Anand, Ünal Sakoğlu · Procedia Computer Science · 2025
Thyroid disease is a widespread and potentially serious medical condition affecting millions of people worldwide. Accurate diagnosis and classification of thyroid disorders is crucial for proper treatment and management of the disease. This work aimed to develop a robust machine learning model for classifying different thyroid conditions based on a comprehensive thyroid dataset. The dataset contained various features and parameters related to thyroid function, enabling the classification of patients into seven distinct thyroid condition categories. Three state-of-the-art machine learning algorithms were employed: Random Forest, Gradient Boosting, and Decision Tree. These ensemble methods are known for their strong predictive capabilities and ability to handle complex, non-linear relationships within the data. The models were rigorously trained and evaluated using appropriate techniques, including cross-validation and stratified sampling, to ensure reliable and generalizable results. Performance metrics such as the F2 score were utilized to assess the models’ performance and ability to handle class imbalances effectively. Through extensive experimentation and fine-tuning of hyperparameters, the Gradient Boosting model emerged as the top performer, achieving an impressive F2 score of 0.97. The model was trained using a comprehensive dataset and evaluated through 5-fold cross-validation, achieving an overall accuracy of 94.7%. Precision and recall scores were 93.5% and 92.8%, respectively, demonstrating the model’s strong performance. Comparative analysis with recent studies shows a 3.5% improvement in accuracy, confirming the efficacy of the proposed approach in thyroid disorder classification. Various other performance metrics based on the analyses of the multi-class confusion matrix resulting from the classification tests further elucidated the model’s strengths and weaknesses, providing valuable insights into its classification capabilities across different thyroid conditions. The findings of this work demonstrate the potential of machine learning to aid in thyroid disease diagnosis and classification, ultimately contributing to improved patient care and more informed clinical decision-making.