An Intelligent Thyroid Diagnosis System Utilizing Multiple Ensemble and Explainable Algorithms With Medical Supported Attributes

Abstract

The widespread impact of thyroid disease presents a challenging task for healthcare experts. Conventional methods for diagnosing this vital condition are often complex and time-consuming. A data-driven approach may provide predictive solutions; however, it necessitates the consideration of all relevant attributes, which can be computationally expensive. This study utilizes the Thyroid Disease dataset to develop an Intelligent Thyroid Diagnosis System that employs various ensemble and explainable algorithms to enhance diagnostic accuracy. The system incorporates Random Forest, Decision Tree, Gradient Boosting, and AdaBoost algorithms, alongside advanced ensemble techniques such as a Voting Classifier combining Random Forest, Decision Tree, and Gradient Boosting with soft voting, a Bagging method utilizing Random Forest, and a Stacking method integrating Random Forest, Decision Tree, Gradient Boosting, AdaBoost with LightGBM. Additionally, we explore the 3SHANN model and a Voting Classifier based on Boosted Decision Tree and ExtraTree. Notably, the Voting Classifier (Boosted Decision Tree + ExtraTree) achieved outstanding performance, attaining 100% accuracy, demonstrating the effectiveness of ensemble techniques in enhancing thyroid disease diagnosis.