Abstract
Abstract Diabetes is a serious health concern worldwide, with its increasing effects due to modern dietary habits that often include excessive sugar and fat consumption. Diabetes affects 1 in 9 adults globally, with over 40 % undiagnosed, and is projected to rise to 1 in 8 by 2050 , impacting 853 million people. Performing a timely and accurate prediction of diabetes is essential for prompt intervention and proper management. This study presents a fused machine learning framework for early-stage diabetes prediction using a symptom-based dataset. The dataset consists of 520 instances, including 200 non-diabetic and 320 diabetic cases, providing a balanced representation for model training and evaluation. The proposed approach integrates multiple classifiers, including Support Vector Machine (SVM), Artificial Neural Network (ANN), Random Forest (RF), and Gradient Boosting (GB), combined using a fuzzy logic-based fusion mechanism to enhance prediction robustness. Experimental results demonstrate that the proposed model achieves a testing accuracy of 99.35%, outperforming individual classifiers in terms of accuracy, precision, and recall. The high performance is attributed to the effective combination of multiple models and the strong predictive nature of symptom-based features. The results indicate that the proposed framework is a promising tool for early diabetes detection, with potential applications in clinical decision-support systems.