Table 5.
Summary of reviewed techniques in diabetes.
| Ref. | Base Learner Models | Ensemble Model | Data Type | Preprocessing Technique | Positive/Negative Cases | Dataset | Attributes/Instances | Accuracy | Best Model |
|---|---|---|---|---|---|---|---|---|---|
| [52] | MLP, SVM, DT, LR | RF, Stacking | Outlier detection and elimination, SMOTE Tomek for imbalanced data | 73/330 | Type 2 Diabetes | 19/403 | Bagging (RF) = 92.5% Stacking = 96.7% |
Stacking | |
| [6] | RF, LR, NB | Soft voting classifier, AdaBoost, Bagging, XGBoost, | Clinical | Min–max normalisation, label encoding, handled missing values | 268/500 | Pima Indians Diabetes | 9/768 | Bagging = 74.8%, Boosting(AdaBoost) = 75.3%, Boosting (XGBoost) = 75.7%, Voting = 79.0% |
Voting |
| [53] | SVM, KNN, DT | Bagging, Stacking | Clinical | SMOTE, k-fold cross validation | KFUH Diabetes | 10/897 | Bagging = 94.1%, Stacking = 94.4% | Stacking | |
| [54] | KNN, LR, MLP | AdaBoost, Stacking | Feature selection, handling missing values | 60/330 | Vanderbilt University’s Biostatistics program | 18/390 | Boosting (AdaBoost) = 91.3%, Stacking = 93.2% |
Stacking | |
| [55] | XGB, CGB, SVM, RF, LR | XGBoost, RF, CatBoost | Missing values eliminated, class imbalance handling, feature selection | 33,332/73,656 | NHANES | 18/124,821 | Boosting(XGBoost) = 70.8%, Bagging(RF) = 78.4%, Boosting(CatBoost) = 82.1% |
Boosting | |
| [38] | LR, SVM | RF, XGBoost | Clinical | feature Selection | 5532/15,599 | Cardiovascular disease | 123/21,131 | Bagging (RF) = 85.5%, Boosting (XGBoost) = 86.2% |
Boosting |
| [17] | SVM | Majority voting, stacking | Clinical | Cross-validation | 268/500 | Pima Indians Diabetes | 9/768 | Stacking = 79%, Voting = 65.10% | Stacking |
| [22] | ANN, SVM, KNN, NB | Bagging, RF, Majority Voting | Clinical | 268/500 | Pima Indians Diabetes | 9/768 | Bagging(RF) = 90.97%, Bagging = 89.69%, Voting = 98.60%, |
Voting | |
| [34] | KNN, RF, DT, SVM, MLP, GB | RF, AdaBoost, Stacking | Clinical | Feature selection with genetic algorithm | 268/500 | Pima Indians Diabetes | 9/768 | Bagging (RF) = 93%, Boosting (GBC) = 95%, Stacking = 98.8% |
Stacking |
| [10] | DT, KNN, SVM | Bagging, Boosting, RF | Clinical | Discretisation, resampling, PCA | 268/500 | Pima Indians Diabetes | 9/768 | Bagging (RF) = 89.7%, Bagging = 89.5%, Boosting = 90.1% |
Boosting |