DefinePK

Abstract

Diabetes type prediction is a difficult issue that is receiving more and more attention. Retinal pictures from the Indian Diabetic Retinopathy Image (IDRiD) dataset have been used to demonstrate the weighted ensemble framework for label prediction utilising different machine learning models that can be suggested to improve the prediction of diabetes. This paper proposes a weighted ensemble classifier for diabetic classification and prediction. Several Machine Learning (ML) classifiers, such as k-nearest Neighbour, Decision Trees, Random Forest, AdaBoost, Naive Bayes, and XGBoost, in addition to Multilayer Perceptron (MLP), were used to improve the prediction of diabetes. The matching Area Under ROC Curve (AUC) of the machine learning model is used to compute the weights. K-fold cross-validation, feature selection, data standardisation, filling in missing values, and outlier rejection are all important considerations for developing diabetes prediction. The grid search technique is used to improve the performance metric, AUC, during hyperparameter tweaking. The grid search technique is used to improve the performance metric, AUC, during hyperparameter tweaking. Because the ML model's AUC is independent of the class distribution, it is used as the model's weight for voting ensembling rather than accuracy. According to experimental study, the suggested system performs better than earlier state-of-the-art methods in terms of recall, precision, and f measure.