Effective Classification of Diabetes Mellitus Using Support Vector Machine Algorithm

Abstract

Diabetes mellitus is one of the metabolic disorders that results in blood glucose levels that are higher than normal. Insulin regulates the flow of sugar from the blood into our cells, where it can be stored or utilized as fuel. Diabetes makes the body unable to create enough insulin or to utilize it effectively. Attempts to classify patients with the disease have been made but these attempts mostly fall short of the necessary diagnostic accuracy and reliability. Early diagnosis and proper management of those with the disease play a crucial role in lowering the burden on patients and can significantly reduce diabetes-related complications and enhance their quality of life. This study aims to develop a simple, non-invasive model to classify diabetes mellitus based on readily recognizable and related risk factors using machine learning. The model was developed using six (6) input variables. Missing value and outlier removal, min-max normalization, feature extraction and feature selection using principal component analysis were methods of data preprocessing applied to the raw dataset. 5-fold cross-validation was used to train and validate the model. Performance evaluation measures included accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AU-ROC). The model development and analysis were carried out using Python 3.9.The classification accuracy, precision, F1-score, and recall for the healthy class were 98%, 97%, 98%, and 98%, respectively while for the diabetic class, 99% precision, 97% recall, 98% F1-score, and 98%accuracy were obtained. This shows that the model can be used to reliably classify diabetes mellitus with high accuracy.