A Comparative Performance Analysis of Machine Learning Approaches for the Early Prediction of Diabetes Disease

Diabetes is a chronic condition that affects a large number of people. Diabetes can be predicted early on, which can lead to better treatment. One of the main causes of this metabolic condition is a hormonal imbalance. Insulin, the hormone that regulates blood sugar, is the hormone that is affected. The condition causes the patient's body to either produce insufficient insulin or to utilize it inefficiently and effectively. Every year, the same disease is responsible for the deaths of 1.6 million people. As per the report of WHO i.e, World Health Organization, the number of diabetic patients has being increased day to day due to different parameters namely bacterial or even viral infection, toxic or even chemical contents mixed with food and so on. The aim of this article is to leverage significant features to generate a machine learning (ML) based prediction algorithm and discover the best classifier to get the best results when compared to clinical outputs. Further, this article proposes a diabetes prediction system that uses medical data and six ML algorithms namely Logistic Regression (LR), Naive Bayes (NB),Decision Tree(DT) classifier K-Nearest Neighbor (KNN), support vector classifier and Random Forest(RF) Classifiers to diagnose diabetes, the accuracy of all the models are compared and finally recommended the best model for the early prediction of the diabetes disease.