Evaluation of Kernel and Parameter Effects on the Performance of Classifiers for Differentiated Thyroid Cancer Recurrence

Thyroid cancer causes tens of thousands of deaths every year in the world. Treatment of this disease is possible with today's medical conditions. However, as in other types of cancer, recurrence is possible. The Differentiated Thyroid Cancer Recurrence dataset, which includes data from 383 patients and was obtained from the UCI platform, was used in the study. The aim of this study is to compare the classification performances of machine learning algorithms according to various parameters and Kernel functions in order to predict recurrence in thyroid cancer patients. For this purpose, three machine learning algorithms were used to predict differentiated thyroid cancer recurrence. In the K-Nearest Neighbors (K-NN) algorithm, the most appropriate k parameter was determined by trying different neighbor number (k) values. The Support Vector Machines (SVM) algorithm examined various kernel functions and hyperparameter settings. In addition, model performance was compared by optimizing different tree numbers and depth parameters with the Random Forest algorithm. These methods aimed to determine the model with the highest accuracy and generalization capacity.