Comprehensive framework for thyroid disorder diagnosis: Integrating advanced feature selection, genetic algorithms,...

Area of Science:

• Endocrinology and Medical Informatics
• Application of computational methods in healthcare

Background:

• Thyroid hormones regulate vital physiological functions including metabolism and development.
• Imbalances can lead to severe health issues like cognitive impairment and nervous system damage.
• Conventional diagnostic methods for thyroid disorders are often time-consuming and labor-intensive.

Purpose of the Study:

• To enhance the accuracy and efficiency of diagnosing thyroid disorders.
• To evaluate the performance of various machine learning algorithms.
• To investigate the utility of hybrid Genetic Algorithm (GA)-Machine Learning (ML) models for thyroid disorder detection.

Main Methods:

• Utilized the UCI thyroid dataset for analysis.
• Pre-processed data included handling missing values, feature scaling, and correlation analysis.
• Tested five ML algorithms (LR, RF, SVM, AB, DT) in both default and hybrid GA-ML configurations (GA-RF, GA-LR, GA-SVM, GA-DT, GA-AB).

Main Results:

• Random Forest (RF) achieved the highest accuracy (93.93%) among non-hybrid models.
• The hybrid GA-RF model demonstrated superior performance, reaching an accuracy of 97.21%.
• The GA-RF model excelled across all evaluated metrics, including F1 Score, sensitivity, specificity, precision, and Cohen's Kappa.

Conclusions:

• Hybrid GA-ML approaches show significant potential for improving clinical diagnostic processes.
• The GA-RF model offers a robust and scalable solution for accurate thyroid disorder identification.
• This study highlights the advantages of advanced computational techniques in endocrine diagnostics.