Application of machine learning algorithms to predict the thyroid disease risk: an experimental comparative study.

Area of Science:

• Endocrinology and Medical Informatics
• Application of artificial intelligence in healthcare

Background:

• Thyroid disease affects hormone production and can impact individuals of all ages.
• Diagnosing thyroid disorders via blood tests is challenging due to complex data interpretation.
• Machine learning offers a potential solution for accurate thyroid risk prediction.

Purpose of the Study:

• To compare the performance of eleven machine learning algorithms for predicting thyroid disease risk.
• To identify the most accurate algorithm for forecasting thyroid disorders.

Main Methods:

• Utilized the Sick-euthyroid dataset from the UCI machine learning repository.
• Compared eleven distinct machine learning algorithms.
• Employed F1-score as the primary evaluation metric due to class imbalance, alongside accuracy and recall.

Main Results:

• The Artificial Neural Network (ANN) Classifier demonstrated superior performance.
• The ANN Classifier achieved an F1-score of 0.957.
• This indicates high accuracy in predicting thyroid risk, outperforming other algorithms.

Conclusions:

• Machine learning, particularly ANN Classifiers, shows significant promise in improving the accuracy of thyroid disease risk prediction.
• The F1-score is a crucial metric for evaluating models in imbalanced datasets, common in medical diagnostics.
• This study highlights the potential of AI to aid in the interpretation of complex diagnostic data for thyroid disorders.