AGXLG: Enhancing Hypothyroidism Prediction with Recursive Feature Elimination and Boosting Techniques

Hypothyroidism is a diagnosed illness. It occurs when the thyroid gland fails to produce enough hormones. Triiodothyron (T3) and thyroxine (T4) are two hormones that are essential for regulating metabolism. In this study, the prediction of hypothyroidism using machine learning (ML) is discussed. Ensemble methods are highly effective for a wide range of data types, including numerical, categorical, and image data. The study uses several indicators to evaluate the outcomes. Recall, F1-score, ROC, accuracy, precision, and loss function are a few of them. It ofers several machine-learning methods for identifying and forecasting thyroid problems. Several boosting techniques, including AdaBoost, Gradient Boosting, XGBoost, LightGBM, and CatBoost, are used by the prediction model. The suggested model, known as AGXLG, is a hybrid method based on machine learning. Additionally, it uses Recursive Feature Elimination (RFE) to identify six key traits. RFE selects features based on ranking their importance through recursive elimination. RFE has identified the following six medical indicators as traits from the given dataset for hyperthyroidism such as TSH (Thyroid-Stimulating Hormone) levels, TT3 (Triiodothyronine), TT4 (Thyroxine), FTI, T4U, age. It achieves an impressive accuracy 99.36% during testing. Predictions were made based on a dataset with 30 features across 3,773 cases. This research could also help other researchers find a suitable model for detecting and forecasting hypothyroidism efficiently.