Machine Learning for Predicting Metabolic Syndrome and its Prognostic Factors

Abstract

Abstract This study explores the potential for early detection of Metabolic Syndrome (MetS) using Machine Learning (ML) techniques. Dissection of prognostic components inciting the syndrome could help patients take cautious steps to prevent it in the early stages. MetS outnumbered diabetics by three to one, a 2020 report found that one billion people worldwide were affected. Patients with MetS typically have no symptoms or signs of the condition and are left undiagnosed. These conditions include extensive circulatory tension, high glucose levels, muscular overload around the abdomen, and unusual levels of cholesterol or fat. Supervised ML techniques like Naïve Bayes, Support Vector Machines, Random Forest, Logistic Regression, C4.5, Cart, etc. are widely used for predictions and diagnoses in various fields. It has been extensively used in medical sciences as well. ML is used for the prediction of the progression of certain diseases and analysis of important parameters in the medical domain. This research uses the aforementioned algorithms for the prediction of MetS using the patients’ dataset. The results were analyzed using precision-recall and Area Under the Curve (AUC) of Receiver Operating characteristic Curve (ROC). The results showed that Naïve Bayes predicted MetS more accurately showing 94.1% accuracy than the rest of the algorithms, while Random Forest surpassed the other tree-based algorithm. According to the results of this research, the prognosis factors for MetS identification are hyperglycemia, dyslipidemia, or a combination of high-density lipoprotein (HDL) dyslipidemia, hypertension, and obesity. Monitoring these factors reduces the risk of MetS occurrence, assists in the prevention, and provides important information for treatment in the early stages.