Explainable AI Based Machine Learning Models for Thyroid Disease Prediction

Abstract

The incidence of thyroid disorders is among the most widespread endocrine diseases worldwide and accurate and timely diagnosis is an important part in patient management/procedures. The paper presents a proposal of an Explainable Artificial Intelligence (XAI) based solution for thyroid disease prediction which predicts using the different machine learning methods including Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, CatBoost, and XGBoost. To address this challenge of class imbalance, pre-processing was carefully done to this data set in order to alter this issue as well as increase the representation as feature, and the performance has been evaluated in terms of precision, recall, accuracy, and F1-score. In comparison to all the models, XGBoost turned out to be the best model with accuracy 98.5% upholding classification of Hyperthyroid, Hypothyroid and Negative thyroid accurately. In order to help ensure clarity and confidence regarding the patient's clinical condition, a method that used SHAP XAI was utilized and interpretation of the trained model was derived into how and why the models arrived at such decisions, thus giving insight into features. This includes XAI which can be used to turn the predictive power into an explainable system to ensure that health practitioners understand the process of how powerful decisions are being made through a scoring rubric. All in all, this study demonstrates that the fusion of XGBoost with XAI techniques is an effective, correct and explainable method to detect thyroid disorders at the very earliest stage, which further assists in enforcing better diagnostic practices that improve patient outcomes.