An Improvised Diabetes Prediction System Using Hybrid Ensemble Approach

High prevalence and substantial health implication of diabetes makes diabetes prediction an important area in medical research. By allowing for prompt actions, early and precise prediction can greatly enhance patient outcomes. We built an ensemble model that merges Support Vector Regression (SVR) and Logistic Regression (LR) to enhance predictive accuracy. The dataset is extensive, containing 2,000 patient records. By using both methods in tandem, we may take advantage of SVR's robustness against non-linear interactions and LR's clarity and ease of interpretations. The predictive power and model complexity are both adequately addressed by this integrated methodology. Our model was successfully trained to generalize the unseen data using Linear Discriminant Analysis (LDA), which helped us avoid overfitting. The results of this study show that the ensemble model outperformed the standalone models, reaching an accuracy of 98.37%. The model demonstrated strong performance in other important parameters as well, with an F1 Score of 97.52%, Precision of 97.48%, and Recall of 98.37%.