Machine Learning Techniques for the Classification of Thyroid Disease

Abstract

In the healthcare system, dealing with a large amount of data is challenging. The techniques of machine learning are used in dealing with such data. As per NFHS – 5 statistics, thyroid diseases are increasing in India. Roughly 1 in 10 Indian adults suffer from a thyroid disorder. It has been expected that more than 42 million peoples suffer from thyroid disease. For the proper diagnosis of disease, it is vital to process the medical data accurately. This study classified thyroid disease cases into hyperthyroid, euthyroid, hypothyroid, and sick. This paper aims to inspect Logistic regression for multiclass categorizing the thyroid dataset. This logistic regression model is evaluated based on its precision, recall, F measure, ROC, RMS Error, and accuracy metrics. Based on the thyroid dataset, we find that logistic regression using the One-vs.-Rest heuristic is 85% accurate, while logistic regression using the multinomial is 86% accurate.