Applications of Thyroid Disease Detection Using Machine Learning

The research seeks to offer valuable insights and references for the prediction of thyroid illness, with a particular focus on accuracy assessment. Through a comprehensive examination, it evaluates the efficacy of three commonly utilized machine learning (ML) algorithms: K-Nearest Neighbors (KNN), Decision Trees, and Logistic Regression. This analysis not only investigates their performance but also elucidates their utilization as classification tools within the context of thyroid disease prediction. Emphasizing the simplicity and effectiveness of logistic regression, especially in binary outcome prediction tasks, the abstract underscores its significance. Decision trees are described as versatile instruments with a hierarchical decision-making structure adaptable for various classification and regression tasks. In contrast, KNN simplifies supervised learning by categorizing new instances based on their similarity to existing ones, ensuring consistency in predictions. In my research, I have developed and applied three distinct algorithms. Notably, the decision tree algorithm exhibited the most promising performance, achieving an accuracy rate of 99 %. Additionally, the logistic regression model demonstrated an accuracy of 88%, while the K-nearest neighbors (KNN) approach yielded a respectable accuracy of 93 %. Leveraging the thyroid dataset from the UC Irvine Finding Knowledge in Databases Archive, this research aims to advance the understanding of thyroid disease prediction methodologies while contributing to improved assessment practices. By providing insights into these ML techniques, the study endeavors to enhance the accuracy and effectiveness of thyroid illness prediction models, ultimately benefiting clinical decision-making and patient care.