Machine Learning-Based Classification of Thyroid Disease: A Comprehensive Study on Early Detection and Risk Factor Analysis

Thyroid disease (TD) arises when the thyroid fails to produce a sufficient quantity of thyroid hormones or when an abnormal growth in the thyroid gland leads to the formation of a lump or nodule. Detecting this condition early on is crucial for preventing or minimizing its impact. This study focuses on creating an efficient framework for classifying TD using different machine learning techniques and wrapper and embedded feature selection methods. Additionally, significant risk factors associated with TD are identified within this proposed system. The dataset utilized in this research is sourced from the University of California Irvine (UCI) repository. Feature selection is performed using forward selection, backward elimination, LASSO and Recursive Feature Elimination technique. Different ML algorithms as K Nearest Neighbor, Support Vector Machine, Stochastic Gradient Descent, Naïve Bayes and logistic Regression are employed to train the mode, and an 8 fold cross validations conducted o assess their performance. To compare the effectiveness of the algorithm various performance metrics are examined. The proposed system is anticipated to be valuable for physicians and patients in the classification of TD and gaining insights into associated risk factors.