Thyroid Disease Multi-class Classification based on Optimized Gradient Boosting Model

Abstract

Human healthcare is one of the most important issues in society to ensure that patients receive the care they require as quickly as possible. One of the disorders that affects the global population and is becoming more prevalent is thyroid disease. Medical information systems are crucial in their capacity to diagnose thyroid disease. Artificial intelligence has recently offered fresh approaches to the existing clinical treatment issues and has demonstrated promising results for individualized diagnosis and therapy planning. Hence, this paper proposes an optimized multi-class classification model, which depends on XGBoost to classify patients with different types of thyroid disease. The main contribution is to (i) propose a Multiclass-Classification for the purpose of diagnosing three different thyroid diseases, (ii) raise the row dataset's feature selection accuracy for classification. (iii) utilize the highly selective XGBoost algorithm for the chosen characteristics, (iv) show that The XGBoost has the best performance and recall, making it the top choice for data analysis in terms of classifying thyroid disease, and (v) improve upon findings from earlier studies by doing the proposed study. XGBoost is trained and tested using UCI machine learning repository dataset of thyroid disease. In addition to build the model with the optimized hyperparameters to achieve and compare the gained results aiming to get the best score of accuracy. From the results, it is shown that the optimized XGBoost achieved 99% accuracy as a win over performing compared with the state of arts models.