A COMPARATIVE ANALYSIS OF MACHINE LEARNING AND DEEP LEARNING-BASED PREDICTIVE FRAMEWORKS FOR DIABETES CLASSIFICATION USING CLINICAL DATA

Abstract

Diabetes mellitus is a major chronic metabolic disorder that requires early identification to reduce the risk of severe health complications. The availability of clinical datasets and advancements in artificial intelligence have provided new opportunities for developing computational approaches for disease prediction. Machine learning (ML) algorithms have been widely investigated for analyzing clinical parameters and assisting healthcare decision-making. In this study, a comparative analysis of conventional machine learning and deep learning-based predictive frameworks was performed for diabetes classification using clinical data. The Pima Indians Diabetes Dataset was utilized, containing important clinical attributes associated with diabetes risk, including glucose level, body mass index, insulin concentration, age, and hereditary factors. Data preprocessing and balancing techniques were applied to improve model performance. Multiple machine learning classifiers, including Random Forest, Support Vector Machine, Decision Tree, and Extra Trees, were evaluated and compared with a deep learning-based predictive model. The performance of each model was assessed using accuracy, sensitivity, specificity, Matthews Correlation Coefficient (MCC), and Area Under the Curve (AUC). The comparative analysis demonstrated that deep learning approaches provided improved predictive capability by automatically learning complex relationships among clinical features, whereas traditional machine learning models showed effective but comparatively limited classification performance. The findings highlight the potential of artificial intelligence-based frameworks for supporting early diabetes screening and personalized healthcare applications.