غزاله کاکاوند تیموری

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.

Diabetes, a metabolic disorder, poses significant annual risks due to various factors, requiring effective management strategies to prevent life-threatening complications. Classified into Type 1, Type 2, and Gestational diabetes, its impact spans diverse demographics, with Type 2 diabetes being particularly concerning due to cellular insulin deficiencies. Early prediction is crucial for intervention and complication prevention. While machine learning and artificial intelligence show promise in predictive modeling for diabetes, challenges in interpreting models hinder widespread adoption among physicians and patients. The complexity of these models often raises doubts about their reliability and practical utility in clinical settings. Addressing interpretability challenges is crucial to fully harnessing predictive analytics in diabetes management, leading to improved patient outcomes and reduced healthcare burdens. Previous research has utilized various algorithms like Naïve Bayes, Support Vector Machines (SVM), K-Nearest Neighbors (KNN), and decision trees for patient classification. In this study using the Pima dataset, we applied a preprocessing technique that utilized the most important features identified by the Random Forest algorithm and we used an ensemble method combining the SVM algorithm and Naïve Bayes for the model. In the first section of the proposed method, we provided explanations regarding the dataset. In the second section, we elucidated all preprocessing steps applied to this dataset, and in the third section, we evaluated the model using the selected algorithm under investigation. The proposed model, after going through the various stages, was able to report an accuracy of 81.82%, a precision of 82.34%, an AUC of 88.19% and a Recall of 70.68%. Considering the review of similar studies, an improvement of 3.99% in accuracy demonstrates a significant advancement that highlights the benefits of traditional methods in disease prediction. These findings suggest the potential use of web-based applications to encourage both physicians and patients in diabetes prediction efforts.