Optimizing Binary Classification Algorithms Through Hyperparameter Tuning: A Case Study on Thyroid Cancer Prediction

The early and accurate detection of thyroid cancer is crucial for informing clinical decisions and improving patient outcomes. Fine-tuning binary classification algorithms through hyperparameter adjustments is essential for improving the performance and predictive accuracy of machine learning models in diagnostic processes. This study examines the effect of hyperparameter tuning on the performance of binary classification algorithms using a real-world dataset of thyroid cancer, comprising ultrasound, cytology, and blood test data. This investigation is a comprehensive case study on enhancing various binary classification algorithms in predicting thyroid cancer, a vital component of medical diagnostics. We apply three tuning methods — grid search and random search — in conjunction with three commonly utilized classification algorithms: Random Forest, Support Vector Machines (SVM), and Logistic Regression. Model performance is assessed using Accuracy, ROC-AUC Score, Confusion Matrix, and ROC Curves before and following tuning. The findings demonstrate that the optimized models significantly outperform their default counterparts, with hyperparameter tuning influencing the rankings of model selection. These results underscore the need to integrate hyperparameter optimization into the machine learning workflow, particularly in critical domains such as medical diagnostics. Moreover, we investigate the impact of data preprocessing, feature selection, and cross-validation techniques on the model outcomes.