Exploring an Efficient Machine Learning Technique for Diabetes Mellitus Prediction

Abstract

Diabetes Mellitus is the most common disease characterized by hyperglycemia resulting from a deficit in the production or action of insulin. Untreated diabetes can lead to a myriad of complications. In 2021, it was estimated that 537 million individuals have diabetes, which is expected to increase to over 783 million by 2045. Preventing and managing this epidemic poses a significant challenge owing to various issues and barriers such as inadequate access to healthcare and lack of surveillance data. However, the advent of machine-learning (ML) techniques can easily address these critical problems. The main goal of this project is to develop a model capable of accurately predicting diabetes in patients. Five machine-learning classification algorithms were employed in this analysis to detect diabetes. The experiments were performed using a diabetes prediction dataset sourced from Kaggle. The efficiency of all five algorithms was assessed using various metrics such as the Kappa statistic, F-measure, precision, true positive rate, and recall. Accuracy was evaluated based on correctly classified and incorrectly classified instances. The results specify that the bagging algorithm outperformed the various other algorithms with the highest accuracy of 97.16% compared to other different algorithms..