Machine Learning For Non- Invasive Diagnostics Of Glucose Metabolism Disorder

Abstract

Glucose metabolism disorder known as Diabetes Mellitus is a state created by uncontrolled blood sugar that may lead to serious damage to multiple organs in patients. Identifying and predicting this disease will save human life. While designing medical diagnosis software, disease prediction is said to be one of the capricious tasks. In the current scenario, many researchers have provided their ideas on using machine learning and artificial intelligence for automated prediction of Diabetes Mellitus. A set of five popular Naïve Bayes, Random Forest, SVM, KNN and Decision Tree have been identified as well as a set of four rarely used GPC, QDA, LDA and AdaBoost have been identified from literature survey. The study is an effort to make a comparative report of the accuracy of two sets and identify the best performer. In conclusion, Support Vector Machine achieved highest accuracy with 81.00% in popular classifiers whereas Linear Discriminant Analysis achieved highest accuracy with 82.00% in less frequently used classifiers. Hence, more such rarely used classifiers should be explored for the realistic health management of diabetes.