Evaluation of Classification Models for Predicting Mortality Rate Using Thyroid Cancer Data

Abstract

Machine Learning (ML) can potentially enhance predictions in real-life domains. This study presents an evaluation and comparison of different ML methods which can be applied on thyroid cancer dataset, called Prostate, Lung, Colorectal and Ovarian (PLCO), of approximately 155,000 participants with thyroid cancer occurrence and mortality incidence. The ML models are explored for predicting mortality rates of patients with thyroid cancer. These models include the Logistic Regression model (LR), K-Neighbors model (KN), Support Vector Classifier (SVC), Gaussian Naïve Bayes (GNB), decision tree classifier (DT), Random Forest classifier (RF), ada boost classifier (AdaB) and Gradient Boosting classifier (GB). The results reveal that AdaB and GB classifiers have the best performance among the models. The results also show that different predictive models can significantly differ with others in terms of their performance evaluated by various metrics. This study shows that the chosen parameters for classifiers will affect their performance; therefore, it is important to explore and evaluate them before final implementation.