Diabetes Prediction using Machine Learning and Explainable Artificial Intelligence Techniques

Abstract

Diabetes influences 537 million human beings globally and may result in diverse fitness issues, inclusive of coronary heart disease, kidney disease, nerve damage, and diabetic retinopathy. A new diabetes forecast framework has evolved with a non-public Bangladeshi dataset and diverse AI methods. The version makes use of a semi-controlled version with excessive inclination aid to expect insulin tires and makes use of algorithms like Decision Tree, SVM, Random Forest, logistic regression, KNN, and different organization methods. After getting ready and checking out all the older models, the proposed framework gave satisfactory results inside the XGBoost classifier with the ADASYN method with 80% accuracy,0.81 F1 coefficient, and an AUC of 0.84, with a 99.3% accuracy completed the use of a mixture of 3 classifiers (Stack). The version additionally makes use of area variant strategies to illustrate its flexibility. The source code is publicly accessible at https://github.com/diabetes_prediction.