<![CDATA[Early detection of diabetes mellitus is a crucial step in preventing chronic complications and enabling more effective disease management. This study aims to analyze the impact of the Recursive Feature Elimination (RFE) method on the performance of machine learning-based diabetes prediction models at RS PKU Muhammadiyah Bima. A quantitative approach was employed by implementing the CRISP-DM framework, encompassing data selection, preprocessing, transformation, data mining, and model evaluation. Missing values in height and weight variables were imputed using linear regression based on age and gender features. The transformation process included calculating the Body Mass Index (BMI) as a new feature relevant to diabetes risk. Evaluation was carried out on three classification algorithms—Support Vector Machine (SVM), Logistic Regression (LR), and Random Forest (RF)—both before and after the application of RFE. The results showed that all models experienced significant performance improvements following feature selection with RFE, achieving 100% in all evaluation metrics. Insulin and BMI were consistently selected features, underscoring their contribution to diabetes detection. It can be concluded that RFE effectively reduces model complexity without sacrificing accuracy, thereby supporting the efficient implementation of predictive models in clinical settings.]]>
Copyrights © 2025
