Comparative Evaluation of Classification Techniques for Predicting Risk and Recurrene of Thyroid Disorders

This article compares various classification techniques in their ability to predict risk levels and recurrence in patients with thyroid disorders. Focusing on a detailed dataset incorporating clinical and pathological features, four prominent classification methods were implemented and evaluated: Logistic Regression, Decision Trees, Random Forests, and Support Vector Machines (SVM). The findings revealed that Random Forests achieved the highest accuracy in predicting the risk level (88.70%) and the determination of disease recurrence (96.52%), outperforming the other evaluated techniques. Logistic Regression and Decision Trees also demonstrated solid performance, with accuracies exceeding 80% for both target variables, while SVM exhibited comparatively lower performance. This analysis highlights the importance of carefully selecting classification algorithms based on specific prediction objectives in medical studies, underscoring the effectiveness of Random Forests for this dataset. The research significantly contributes to the field of medical analytics, offering critical insights for the development of more accurate and efficient predictive models in the diagnosis and monitoring of thyroid disorders.