Background: Class imbalance and limited interpretability remain major barriers to the clinical adoption of machine learning in diabetes prediction. These challenges often result in poor sensitivity to high-risk cases and reduced trust in AI-based decision support. This study addresses these limitations by integrating SMOTE-based resampling with SHAP-driven explainability, aiming to enhance both predictive performance and clinical transparency for real-world deployment. Objective: To develop and validate an interpretable machine learning framework that addresses class imbalance through advanced resampling techniques while providing clinically meaningful explanations for enhanced decision support. This study serves as a methodologically rigorous proof-of-concept, prioritizing analytical integrity over scale. While based on a computationally feasible subset of 1500 records, future work will extend to the full 100,000-patient dataset to evaluate scalability and external validity. We used the publicly available, de-identified Diabetes Prediction Dataset hosted on Kaggle, which is synthetic/derivative and not a clinically curated cohort. Accordingly, this study is framed as a methodological proof-of-concept rather than a clinically generalizable evaluation. Methods: We implemented a robust seven-stage pipeline integrating the Synthetic Minority Oversampling Technique (SMOTE) with SHapley Additive exPlanations (SHAP) to enhance model interpretability and address class imbalance. Five machine learning algorithms—Random Forest, Gradient Boosting, Support Vector Machine (SVM), Logistic Regression, and XGBoost—were comparatively evaluated on a stratified random sample of 1500 patient records drawn from the publicly available Diabetes Prediction Dataset (n = 100,000) hosted on Kaggle. To ensure methodological rigor and prevent data leakage, all preprocessing steps—including SMOTE application—were performed within the training folds of a 5-fold stratified cross-validation framework, preserving the original class distribution in each fold. Model performance was assessed using accuracy, area under the receiver operating characteristic curve (AUC), sensitivity, specificity, F1-score, and precision. Statistical significance was determined using McNemar’s test, with p-values adjusted via the Bonferroni correction to control for multiple comparisons. Results: The Random Forest-SMOTE model achieved superior performance with 96.91% accuracy (95% CI: 95.4–98.2%), AUC of 0.998, sensitivity of 99.5%, and specificity of 97.3%, significantly outperforming recent benchmarks (p < 0.001). SHAP analysis identified glucose (SHAP value: 2.34) and BMI (SHAP value: 1.87) as primary predictors, demonstrating strong clinical concordance. Feature interaction analysis revealed synergistic effects between glucose and BMI, providing actionable insights for personalized intervention strategies. Conclusions: Despite promising results, further validation of the proposed framework is required prior to any clinical deployment. At this stage, the study should be regarded as a methodological proof-of-concept rather than a clinically generalizable evaluation. Our framework successfully bridges algorithmic performance and clinical applicability. It achieved high cross-validated performance on a publicly available Kaggle dataset, with Random Forest reaching 96.9% accuracy and 0.998 AUC. These results are dataset-specific and should not be interpreted as clinical performance. External, prospective validation in real-world cohorts is required prior to any consideration of clinical deployment, particularly for personalized risk assessment in healthcare systems.

Accurate used-vehicle price prediction is essential for consumers, dealers, and financial institutions, as pricing dynamics involve complex and non-linear relationships influenced by vehicle condition, depreciation patterns, and heterogeneous market factors. While ensemble learning models have demonstrated strong predictive capabilities, existing studies rarely compare them systematically with linear regression under multiple data partitioning strategies. This study proposes an Optimized Data Partitioning Framework (ODPF) to evaluate model performance and stability across four train–test split ratios (50–50, 60–40, 70–30, and 80–20) using a leakage-free preprocessing pipeline. The framework incorporates a variance-based stability index to quantify the effect of sampling variability, a methodological dimension largely absent from prior vehicle-pricing research. Six algorithms (Linear Regression, Decision Tree, Support Vector Regression (SVR), Random Forest, XGBoost, and LightGBM) were evaluated under consistent preprocessing and experimental conditions. The results indicate that ensemble methods outperform Linear Regression across all evaluation metrics, and Random Forest demonstrated the strongest performance, with a Root Mean Square Error (RMSE) and a coefficient of determination (R²) of 274.26 and 0.9995, respectively. XGBoost and LightGBM also exhibited high accuracy (R² > 0.998), whereas SVR showed limited generalization (RMSE = 1232.73; R² = 0.0800) for the sales analytics dataset. Stability analysis across repeated sampling identifies the 80–20 split as the most reliable configuration, exhibiting lower performance variance and stronger generalization consistency. Overall, the findings indicate that algorithm selection has a greater influence on predictive accuracy than the partition ratio alone, providing practical guidance for developing robust pricing models in the automotive domain.

Cardiovascular Disease (CVD) remains the leading cause of death worldwide, accounting for over 17.9 million deaths annually. Traditional risk assessment tools such as the Framingham Risk Score and Atherosclerotic Cardiovascular Disease (ASCVD) calculator are constrained by linear assumptions and limited variables, often failing to capture complex interactions among clinical and behavioral factors. To overcome these limitations, this study proposes an Adaptive Risk-Stratified Stacking (ARSS) framework that integrates ensemble learning, Explainable Artificial Intelligence (XAI), and Bayesian uncertainty quantification for ten-year CVD prediction. Using data from the Framingham Heart Study (FHS) (n = 4,240; 16 features), the framework combines Random Forest, Extreme Gradient Boosting (XGBoost), and Logistic Regression as base learners, with a Logistic Regression meta-classifier trained using five-fold stratified cross-validation. The adaptive stratification mechanism enables subgroup-specific learning across low-, intermediate-, and high-risk cohorts, enhancing personalization and sensitivity. The ARSS model achieved 89.6% accuracy, an F1-score of 0.89, and an area under the receiver operating characteristic curve (ROC–AUC) of 0.918 (95% Confidence Interval (CI): 0.907–0.929), significantly outperforming baseline models (p < 0.01, Cohen's d ≥ 0.71). Calibration analysis indicated strong reliability (Brier Score = 0.076), whereas Shapley Additive Explanations (SHAP)-based interpretability revealed clinically consistent feature interactions such as Age × Systolic Blood Pressure and Diabetes × Glucose, reinforcing the model's physiological plausibility. Bayesian uncertainty estimation further enhanced confidence in predictive reliability and transparency. Overall, the proposed ARSS framework demonstrates that interpretable, risk-stratified ensemble learning can bridge predictive accuracy with clinical trustworthiness, establishing a unified and ethical paradigm for XAI in precision cardiovascular prevention.