Enhanced Prediction of Thyroid Disease Using Machine Learning Method

Thyroid disease is becoming increasingly in men, women and children but commonly occurring among women over the age of 30. It causes heart problem, eye problem, fertility and pregnancy problems over its effect for long time. As a result, it is critical to evaluate the thyroid information in order to forecast the early prediction of disease and take steps to avoid the deadly condition of thyroid cancer. This study is based upon designing a model for timely detection of thyroid disease by observing the features from thyroid disease dataset which was accessed from UCI repository site by using machine learning algorithms. We have used three machine learning models such as K-Nearest Neighbors (K- NN), decision tree (DT) and multilayer perceptron (MLP) for prediction of thyroid disease and measure the performance of these models in form of accuracy and area under the curve. Comparative analysis of these three models reveals that MLP performs better in classifying thyroid disease with an accuracy value of 95.73 and Area Under the curve with value of 94.23. The planned experiment was carried out on 3163 cases and 24 thyroid characteristics.