Thyro-AI: Harnessing Machine Learning for Thyroid Prediction

Abstract

Thyroid issues are among the common diseases in developing and developed countries, therefore early detection with accurate results is essential in managing the condition effectively. In this study, we proposed a thyroid detection model through a comparative analysis of different machine learning algorithms like Decision tree (DT), support vector machine (SVM), and logistic regression (LR) based on thyroid-related features such as thyroid stimulating hormone (TSH), triiodothyronine (T3) and thyroxine (TT4). We examined the performance of our models based on Accuracy, F1 score, Recall, and AUC with a public thyroid dataset from Kaggle. The DT model outperformed the SVM and LR with an AUC of $\mathbf{9 3. 1 9 \%}$, an F1-score of 88.89%, and an accuracy of $\mathbf{9 8. 6 5 \%}$. The accuracy, recall, and F1-score of the SVM and LR were marginally less than the DT. The results emphasize how the diagnosis can be improved by using machine learning techniques of thyroid ailment, despite admitting the small flaws, which include the small dataset and possibly biased preprocessing methods. This study demonstrates the promise of machine learning in supporting thyroid diagnosis. With further development and validation, these models could become valuable tools for healthcare professionals, potentially leading to earlier diagnoses and improved patient care.

Ähnliche Arbeiten