Comparative Study of Machine Learning Based Diabetes Predictive System

Abstract

Diabetes is one of the most lethal diseases in the world. It is also a precursor to various other disorders such as coronary failure, blindness, and kidney diseases. Patients often need to visit diagnostic centers to get their reports after consultation, which requires a significant investment of time and money. However, with the growth of machine learning methods, we now have the ability to address this issue. Advanced systems utilizing information processing can forecast whether a patient has diabetes or not. Furthermore, early prediction of the disease can provide patients with critical interventions before it fully develops. Data mining techniques can extract hidden information from large datasets of diabetes-related information. The aim of this research is to develop a system that can predict the diabetic risk level of a patient with higher accuracy. The model development is based on classification methods such as K-Nearest Neighbors, Decision Tree, and Support Vector Machine (SVM) algorithms. For K-Nearest Neighbors, the models achieve an accuracy of 71%, 78% for SVM, and 70% for the Decision Tree algorithm. The outcomes demonstrate a significant accuracy of these methods.