Revolutionizing Diabetes Prediction Models: Leveraging Advanced Feature Reduction and Classification Methods

The global challenge of diabetes demands innovative approaches for early diagnosis and effective management. This paper investigates the integration of advanced dimensionality reduction and feature selection techniques to improve diabetes prediction models. Utilizing Principal Component Analysis (PCA) and the Artificial Bee Colony (ABC) algorithm, the study focuses on optimizing feature selection to enhance model accuracy, computational efficiency, and interpretability. Employing two diverse datasets—the Behavioral Risk Factor Surveillance System (BRFSS) and the Pima Indians Diabetes Dataset—the research evaluates the performance of Random Forest and Logistic Regression models. Results reveal significant improvements in predictive accuracy and computational efficiency, particularly with ABC-selected features, reducing feature dimensionality while maintaining high accuracy. This study underscores the potential of combining PCA and ABC to revolutionize predictive healthcare analytics, offering robust tools for diabetes management and prevention.