A Rule-Generative Predictive Clustering Model For Thyroid Prediction Using Learning Approaches

Abstract

Around the world, thyroid disease is considered one of the most highly detected endocrinopathies; for the overall health of a human, thyroid illness is regarded as a significant concern since the thyroid gland controls the human body's metabolism. A dependable, automatic and precise machine-learning (ML) system for thyroid detection is necessary to save time and lower mistake rates. The proposed strategy seeks to overcome previous work's constraints, such as improper comprehensive feature analysis, prediction accuracy improvement, dependability and visualization. Here, 29 clinical factors from a public dataset on thyroid disorders from the California University, Irvine ML repository were employed. By examining symptoms in the early stages and displacing the manual examination of these characteristics, the medical features enabled us to develop a Machine Learning (ML) model that may forecast thyroid-related diseases. Understanding the purpose of features in the prediction tasks of the thyroid is made more accessible with visualization and feature analysis. Furthermore, data balancing and 5-fold CV with using the synthetic minority oversampling technique, intend to solve the over-fitting issue. Because numerous classifiers are involved in the prediction task, learning maintains the trustworthiness of the thyroid forecasting system. Using the proposed Rule-based-Generative Clustering (RGC) with the k-CV method, the suggested model achieved a specificity of 99%, an accuracy of 99.1% and a sensitivity of 99%. This makes it suitable for real-time diagnostic schemes to facilitate disease identification and encourage early-stage treatment.