Enhancing thyroid disease prediction and comorbidity management through advanced machine learning frameworks

Abstract

Thyroid disease is one of the most prevalent endocrine disorders worldwide, necessitating precise and efficient diagnostic models for improved clinical outcomes. This study proposes a Hybrid Feature Selection and Deep Learning Framework (HFSDLF) that integrates Random Forests with Principal Component Analysis (PCA) and L1 regularization for effective feature selection and classification. Utilizing the UCI Thyroid Dataset, the framework combines the strengths of deep learning-based feature extraction and traditional machine learning classifiers. The Random Forest classifier achieved the highest accuracy of 96.30 %, outperforming other models such as Decision Trees and Logistic Regression, with notable improvements in sensitivity and specificity. The novelty of this work lies in its hybrid approach to feature selection, which reduces dimensionality while retaining the most informative features, and its application of an optimized Random Forest model for enhanced classification accuracy. Comparative analysis with existing methods further highlights the superiority of the proposed framework in terms of accuracy and processing efficiency. This research addresses key limitations of existing approaches and contributes to the field by demonstrating a scalable and interpretable solution for thyroid disease diagnosis. The proposed framework provides a benchmark for future studies, underscoring the importance of hybrid methodologies in medical data analysis.