A Robust Ensemble Machine Learning Model using Soft Voting for Thyroid Cancer Prediction

Thyroid cancer (TC), a prevalent malignancy, originates from abnormal cell growth within the thyroid gland. While not universally curable, early diagnosis and appropriate treatment can lead to positive outcomes. Accurate identification of thyroid cancer remains challenging due to the limitations of existing diagnostic techniques in reliably detecting malignant thyroid nodules. This study proposes an expert system named Voting-based Ensemble Intelligent System for Thyroid Cancer Prediction (VEISTCP), leveraging ensemble learning to address these diagnostic limitations. Additionally, VEISTCP employs the SMOTE-ENN hybrid sampling technique to mitigate class imbalance within the thyroid cancer dataset obtained from Sheng Jing Hospital of China. The performance of the proposed model has been evaluated using $\mathbf{1 0}$-fold cross-validation and four standard evaluation metrics: accuracy, precision, recall, and F1-score. A comprehensive comparison has been conducted across various standard machine learning and ensemble learning models, including their voting-based combinations. The experimental results highlight the superior performance of VEISTCP, achieving accuracy of $\mathbf{9 3. 6 \%}$ and F1-score of $\mathbf{9 1. 9 \%}$.