Hybrid Thyroid Stage Prediction Models Combining Classification, Clustering and Ensemble Systems

Abstract

Early and correct detection of thyroid disease is very important for correct and timely treatment. The need to increase the accuracy of detecting and classifying thyroid disease poses a great challenge not only to the research community but also to healthcare industries. Usage of machine learning algorithms for thyroid disease classification is an area of research that is gaining popularity for the past few years. Automatic thyroid disease computer aided system for diagnosing the disease requires sophisticated and effective algorithms to perform classification in an accurate and time efficient manner. As a solution to this demand, hybrid models that combine clustering and classification algorithms along with ensemble technology are proposed. Four category of thyroid disease prediction system are proposed. They are Clustering + Classification models, Classification + Classification Models, Clustering + Clustering Models and Classification + Clustering Models. Two types of ensembles, namely, homogeneous and heterogeneous, are also considered and analyzed. Performance evaluation showed that the Classification + Classification model based on the combination of SVM and heterogeneous KNN + SVM classifier produce highest prediction accuracy.