Background and Objectives: According to the random nature of heuristic algorithms, stability analysis of heuristic ensemble classifiers has particular importance. Methods: The novelty of this paper is using a statistical method consists of Plackett-Burman design, and Taguchi for the first time to specify not only important parameters, but also optimal levels for them. Minitab and Design Expert software programs are utilized to achieve the stability goals of this research. Results: The proposed approach is useful as a Preprocessing method before employing heuristic ensemble classifiers; i. e., first discover optimal levels of important parameters and then apply these parameters to heuristic ensemble classifiers to attain the best results. Another significant difference between this research and previous works related to stability analysis is the definition of the response variable; an average of three criteria of the Pareto front is used as response variable. Finally, to clarify the performance of this method, obtained optimal levels are applied to a typical multi-objective heuristic ensemble classifier, and its results are compared with the results of using empirical values; obtained results indicate improvements in the proposed method. Conclusion: This approach can analyze more parameters with less computational costs in comparison with previous works. This capability is one of the advantages of the proposed method.

Accurate and early prediction of diabetes is crucial for initiating prompt treatment and minimizing the risk of long-term health issues. This study introduces a comprehensive machine learning model aimed at improving diabetes prediction by leveraging two clinical datasets: the PIMA Indians Diabetes Dataset and the Early-Stage Diabetes Dataset. The pipeline tackles common challenges in medical data, such as missing values, class imbalance, and feature relevance, through a series of advanced Preprocessing steps, including class-specific imputation, engineered feature construction, and SMOTETomek resampling. To identify the most informative predictors, a hybrid feature selection strategy is employed, integrating recursive elimination, Random Forest-based importance, and gradient boosting. Model training uses Random Forest and Gradient Boosting classifiers, which are fine-tuned and combined through weighted ensemble averaging to boost predictive performance. The resulting model achieves 93.33% accuracy on the PIMA dataset and 98.44% accuracy on the Early-Stage dataset, outperforming previously reported approaches. To enhance transparency and clinical applicability, both local (LIME) and global (SHAP) explainability methods are applied, highlighting clinically relevant features. Furthermore, probability calibration is performed to ensure that predicted risk scores align with true outcome frequencies, increasing trust in the model’s use for clinical decision support. Overall, the proposed model offers a robust, interpretable, and clinically reliable solution for early-stage diabetes prediction.