Enhanced Diabetes Prediction Through Advanced Machine Learning and Imbalance Handling Techniques

This paper presents a comprehensive analysis of various machine learning classifiers for diabetes prediction, with a specific focus on addressing the challenges posed by imbalanced datasets. The study evaluates the effectiveness of three popular imbalance handling techniques, Adaptive Synthetic (ADASYN), Synthetic Minority Over-sampling Technique (SMOTE), and Random Under-Sampling, when applied to different classifiers, including Gradient Boosting, Decision Tree, K-Nearest Neighbors, and Logistic Regression. Through extensive experimentation on the PIMA Indians Diabetes Dataset and the Diabetes Dataset 2019, the proposed methodology integrates advanced feature engineering and robust imbalance handling to enhance prediction accuracy. The results demonstrate that the Gradient Boosting classifier, particularly when combined with the ADASYN technique, consistently outperforms other models across all evaluation metrics, achieving state-of-the-art performance. This study not only highlights the importance of selecting appropriate imbalance handling methods but also provides practical insights into the strengths and limitations of different classifiers in the context of medical data classification. The findings are further supported by visualizations such as ROC curves, precision-recall curves, and confusion matrices, offering a detailed understanding of model behavior. This research contributes to the development of more accurate and reliable predictive models for diabetes, with potential applications in broader medical data analysis.