<![CDATA[Accurate diabetes classification is a significant challenge in medical diagnostics, especially in imbalanced datasets. This study addresses this issue by introducing A New Modified Weighted SMOTE (ANMWS), integrated with Priority of Attribute by Expert Judgement (PAEJ) framework, to enhance the performance of machine learning models for imbalanced data. PAEJ categorizes attributes into three levels—high, medium and low priority—based on expert knowledge, while ANMWS applies weighted oversampling using these priority levels to generate synthetic data more representative of real-world cases. The proposed method was evaluated using three algorithms: Support Vector Machine (SVM), Logistic Regression, and Naïve Bayes. Results indicate that applying ANMWS algorithm with PAEJ framework significantly improved predictive performance, with AUC values increasing to 0.995 for SVM, 0.993 for Logistic Regression, and 0.990 for Naïve Bayes, compared to 0.980, 0.978, and 0.975, respectively, using standard SMOTE. Additionally, precision and recall for SVM improved by 5% and 7%, respectively. These findings demonstrate the critical role of ANMWS algorithm and PAEJ framework in addressing class imbalance, providing a reliable method for early diabetes diagnosis and informed clinical decision-making.]]>
