Optimized Diabetes Prediction Using SMOTE and Machine Learning Techniques

Diabetes is a long-term illness that affects people all over the world. It can cause serious side effects including as blindness, kidney failure, amputation, heart failure, and stroke. Early detection improves patients' outcomes and care. To boost the data's diversity and robustness, we coupled our unique dataset, Glucosyn, with the Pima Indians Diabetes dataset for this study. Several machine learning approaches were evaluated, including SVM, AdaBoost with Decision Trees, and Bagging. Ensemble approaches and hybrid models increased accuracy, while K-Means clustering improved system knowledge. SMOTE enhanced minority class forecasts by correcting class imbalances. GridSearchCV's hyperparameter adjustments improved performance.