Improving Machine Learning Performance for Diabetes Prediction

Diabetes is a formidable ailment characterized by elevated glucose levels in the bloodstream. Hence, it can cause many complications if it remains untreated and unidentified. Machine learning algorithms play a crucial role in early-stage detection and prediction of diabetes. This research aims to enhance the accuracy of diabetes mellitus prediction by using several techniques within the realm of machine learning. In this regard, the Pima Indians Diabetes dataset was used. Then, different machine learning algorithms have been applied separately or using ensemble learning. As a result, Support Vector Machine achieved an accuracy of 77.89%. Moreover, an empirical assessment of ensemble learning methodology involving the voting classifier and Bagging is conducted alongside contemporary techniques and base classifiers. The ensemble approach demonstrates its prowess, with the voting classifier achieving a result of 90.11%. Notably, the Bagging Classifier emerges as the frontrunner, attaining a remarkable accuracy value of 96.69%.