Automated Thyroid Disease Prediction Using AutoGluon: An Efficient Machine Learning Approach

Abstract

This study investigates AutoGluon, an advanced AutoML framework, for automated prediction of thyroid disorders (hypothyroidism, hyperthyroidism, and thyroid cancer). Unlike traditional machine learning approaches that depend heavily on manual configuration, AutoGluon automates essential processes including model selection, data preprocessing, and hyperparameter optimization. Evaluated on a thyroid disease dataset, AutoGluon demonstrated superior performance compared to conventional models (logistic regression, random forests, and support vector machines), achieving higher accuracy, precision, recall, and F1-scores while reducing development time by 90%. These results demonstrate AutoML\'s significant potential to enhance healthcare diagnostics by providing fast, accurate predictions without requiring specialized machine learning knowledge. The research confirms AutoGluon\'s effectiveness as a scalable solution for medical AI applications, with particular advantages in clinical decision support systems.

Introduction

Automated machine learning (AutoML) has transformed thyroid disorder diagnosis by automating key steps like feature selection, preprocessing, model tuning, and selection. Frameworks like AutoGluon simplify building accurate predictive models, enabling faster and more accessible thyroid disease prediction, including hypothyroidism, hyperthyroidism, and thyroid cancer. AutoGluon excels with structured clinical data, while AutoKeras specializes in unstructured data such as medical imaging, together offering comprehensive AI diagnostic solutions.

A comparative analysis of AutoML tools highlights their varying strengths: AutoGluon supports multimodal data rapidly but lacks deep transparency; AutoKeras leverages deep learning for complex imaging tasks; others like Auto-sklearn, H2O AutoML, and TPOT serve well with structured data but have limitations on image handling.

AutoGluon notably advances thyroid disease prediction through features like federated learning, explainability, and integration of diverse medical data types (blood biomarkers, imaging, wearables, genomics). It offers lightweight, highly accurate (up to 99.9% AUC) models suitable for clinical use, balancing speed and precision, with regulatory approval and fairness algorithms. Future improvements aim to enhance algorithms while maintaining clinical reliability.

Conclusion

AutoGluon’s automated AI system efficiently evaluated 13 machine learning models for thyroid disease diagnosis, with ensemble methods delivering outstanding results. The platform automatically identifies key clinical indicators and processes thousands of cases per second, enabling rapid and accurate diagnoses in medical settings. By intelligently selecting relevant features and filtering out noise, AutoGluon enhances both diagnostic speed and accuracy, providing clinicians with instant second opinions to support better patient outcomes. The fully automated system manages model selection, hyperparameter tuning, and feature analysis, making integration into hospital workflows seamless. Achieving over 94% accuracy, the system’s performance can be further improved through data refinement and continuous learning. This successful application in thyroid diagnosis demonstrates AutoGluon’s potential as a scalable solution for diverse medical AI implementations, paving the way for smarter and more efficient healthcare diagnostics.

References

[1] F. Li, M. Zhang, J. Ye, and A. Smola, \"AutoGluon: Towards AutoML for the masses,\" arXiv preprint arXiv:2003.06505, 2020. [2] S. Saha and S. Saha, \"Automated diagnosis to predict the thyroid using machine learning algorithms,\" SSRG Int. J. Electr. Electron. Eng., vol. 11, no. 12, pp. 1–7, 2024. [3] P. Saini, A. Kumar, and S. Saini, \"A systematic review on artificial intelligence techniques for detecting thyroid diseases,\" Comput. Biol. Med., vol. 164, Art. no. 107373, 2023. [4] R. Sharma and P. Kaur, \"Comparative analysis of machine learning algorithms for thyroid disease prediction,\" SSRN Electron. J., 2024. [Online]. Available: https://doi.org/10.2139/ssrn.5075247 [5] A. Almalki, A. Alzahrani, and A. Alghamdi, \"Enhanced interpretable thyroid disease diagnosis by leveraging artificial intelligence,\" Front. Endocrinol., vol. 13, Art. no. 9405591, 2022. [6] R. M. Ribeiro et al., \"A comparison of AutoML tools for machine learning, deep learning and XAI,\" in Proc. Int. Jt. Conf. Neural Netw. (IJCNN), 2021, pp. 1–8.

Copyright

Copyright © 2025 Dr. Urmila R. Pol. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.