Enhanced diagnosis of thyroid diseases through advanced machine learning methodologies

Thyroid disease is a health concern related to the thyroid gland, which is vital for controlling the metabolism of the human body. Predominantly affecting women in their fourth or fifth decades of life, thyroid disease can result in physical and mental issues. This research focuses on improving the diagnostic process by creating a classification model that utilises various machine learning models and a deeplearning model to categorise three types of thyroid disease conditions. This research developed an automated system capable of classifying three thyroid conditions using five machine learning models and a deep learning model. Resampling techniques, such as SMOTE oversampling and Random undersampling, are utilised to correct the issue of class imbalance in the dataset. Finally, a web-based application is developed utilising the most effective model, GBC, which facilitates easy classification of thyroid diseases. The experimental analysis showed that the Gradient Boosting Classifier (GBC), using oversampling techniques, achieved the highest level of performance in classifying thyroid diseases, obtaining an accuracy and F1-Score of 99.76%. This study demonstrated that TSH was the most indicative biomarker for thyroid disease classification. The experimental results proved that the Gradient Boosting Classifier (GBC) utilising the oversampling technique achieved a superior performance compared to other classifier models, with an accuracy and F1-Score of 99.76%. This research presented insights that can assist healthcare practitioners in promptly diagnosing thyroid diseases.