AI-DRIVEN DIAGNOSIS AND CLASSIFICATION OF DIABETES MELLITUS USING MACHINE LEARNING APPROACHES

Abstract

Recent studies have highlighted diabetes as a chronic disease that is spreading worldwide. The World Health Organization (WHO) reports 422 million patients with diabetes worldwide, and this number will increase if diabetes is not adequately controlled. Worldwide, accurate and early diagnosis of diabetes is required. However, there are still deficiencies in the diagnosis, analysis of features, and classification of diabetes types. The proposed model is based on machine learning approaches, where analysis of the features highlights that glucose is the key factor in diagnosing diabetes and insulin is the main feature used to classify diabetes types. The proposed model consists of two major components. The first part discusses model development and training, which involves pre-processing the dataset, extracting features, and training the model on the PIMA Diabetes dataset. This study used four machine learning classifiers for model training: K-Nearest Neighbors, Logistic Regression, Gaussian Naïve Bayes, and Support Vector Machine (SVM). In the second part, the proposed model is evaluated using the PIMA diabetes open-source dataset. After the evolution of the proposed model, the best accuracy was obtained for the training model at 0.9541. The testing achieved an accuracy of 0.9607, which proves that the proposed model performs exceptionally well.