Date of Publication :8th August 2024

Abstract:In the realm of critical healthcare challenges, Diabetes Mellitus stands as a formidable adversary, affecting a multitude of individuals across the globe. This silent, insidious condition is fuelled by a complex interplay of factors, including age, obesity, sedentary lifestyles, hereditary predispositions, dietary habits, high blood pressure, and more. The consequences of uncontrolled diabetes are dire, encompassing a heightened risk of heart disease, kidney dysfunction, strokes, vision impairments, neuropathies, and a host of other complications. In the ongoing quest to combat this multifaceted malady, the traditional approach within the healthcare domain involves an intricate process of diagnostic testing to pinpoint the nuances of each patient's condition. However, amidst the vast sea of medical data, a revolutionary force has emerged – Big Data Analytics and Machine learning. The power of modern machine learning, blending it with the wisdom of traditional medical factors such as Glucose levels, BMI, Age, and Insulin sensitivity has been harnessed. The aim is to enhance the classification accuracy of diabetes prediction. The existing methods, while valiant, have left room for improvement. Our novel approach, a fusion of sciences, different algorithm promises to be a game-changer. We've introduced a dataset, one that extends the boundaries of what is currently known, and our results speak volumes. As we move forward, we have constructed a seamless pipeline, a roadmap to precision, in the realm of diabetes prediction. Our research, meticulous and unyielding, strives to elevate the accuracy of classification. It is a symphony of data, a harmonious convergence of variables, and a beacon of hope for those grappling with the uncertainties of Diabetes Mellitus. This paper presents a comprehensive study on the application of machine learning techniques for the early prediction and risk assessment of diabetes using the R programming language. We employ advanced machine learning algorithms including Random Forest (RF), Logistical Regression, K-Nearest Neighbors Algorithm and The Naïve Bayes classifier to develop robust predictive models. Our approach integrates feature selection techniques to identify the most influential variables contributing to accurate diabetic prediction. Through rigorous evaluation and comparison, our results demonstrate the superior performance of the proposed models in terms of sensitivity, specificity, and area under the receiver operating characteristic curve (AUC-ROC). The findings herein highlight the potential of machine learning methodologies in revolutionizing early diabetic prediction and underscore the effectiveness of R as a versatile tool for implementing such predictive models. Furthermore, this research investigates the interpretability of the developed models, emphasizing the importance of clinical relevance and transparency in predictive healthcare systems, Through feature selection, we identify crucial variables for precise prediction. Our models outperform existing methods in sensitivity, specificity, and AUC-ROC. This work significantly advances early diabetic prediction and personalized management.

Reference :