Finding Best Voting Classifier for Diabetic Disease Classification

Abstract

Diabetes is a metabolic disorder that affects a large amount of people globally. Among diabetic disorders, diabetes mellitus is the most common ailment. Currently, while treating diabetes, hospitals gather required facts via suitable medical tests and appropriate treatment is administered primarily based on prognosis. Early detection of diabetes is important to prevent diabetes from progressing into a chronic illness. Machine learning helps to identify and predict diabetic even at beginning stage. An ensemble machine learning technique helps to get better classification and higher accuracy to predict from diabetic data set. Different voting classifier techniques are present for ensemble machine learning. The present paper proposes a suitable voting classifier for diabetic classification that can be used to predict diabetes with better accuracy. Analysis of the obtained results, show that hard-voting method provides greater accuracy than soft-voting method, the three diabetes data sets used for the study. Implementation is done by applying ML algorithms on three different diabetes data sets obtained from various repositories. ML-algorithms like Logistic Regression, Ada Boost, Decision Tree, Cat Boost, Naïve Bayes, SVM, XG Boost, RF, KNN, MLP, Bagging Classifier, and Gradient Boost Classifier are used.