Table 1.
Summary of reviewed techniques in Heart Disease.
| Ref. | Base Learner Models | Ensemble Model | Data Type | Preprocessing Technique | Positive/Negative Cases | Dataset | Attributes/Instances | Accuracy | Best Model |
|---|---|---|---|---|---|---|---|---|---|
| [35] | KNN, LR, SVM, RF, CART, LDA | Gradient Boost, RF | Clinical | 139/164 | UCI Cleveland Heart Disease | 14/303 | Bagging (RF) = 83%, Boosting (Gradient) = 81% | Boosting | |
| [4] | LR | RF, AdaBoost, Voting, Stacking | Random oversampling | 644/3594 | Kaggle Chronic Heart Disease | 16/4238 | Bagging (RF) = 96%, AdaBoost = 64%, Voting = 76%, Stacking = 99% | Stacking | |
| [36] | SVM | AdaBoost, Stacking, RF | Clinical | Feature selection | 139/164 | UCI Cleveland Heart Disease | 14/303 | Bagging (RF) = 88.0%, Boosting (AdaBoost) = 88.0%, Stacking = 92.2% | Stacking |
| [37] | SVM | Stacking, RF | Clinical | Feature selection, Optimisation | 139/164 | UCI Cleveland Heart Disease | 14/303 | Stacking = 91.2%, Bagging (RF) = 82.9% | Stacking |
| [19] | XGB, LR, RF, KNN | Majority Voting, XGBoost, RF | Clinical | Feature selection | 139/164 | UCI Cleveland Heart Disease | 14/303 | Voting = 94%, Bagging (RF) = 92%, Boosting (XGBoost) = 87% | Voting |
| [38] | LR, SVM | RF, XGBoost | Clinical | Feature Selection, | 1447/7012 | Cardiovascular disease | 131/8459 | Bagging (RF) = 83.6%, Boosting (XGBoost) = 83.8% | Boosting |
| [39] | XGB, DT, KNN | Stacking, RF, XGB, DT | Eliminating outliers, Scaling | Kaggle Cardiovascular | 13/7000 | Stacking = 86.4%, Bagging (RF) = 88.6%, Boosting (XGBoost) = 88.1%, Bagging (DT) = 86.3% | Bagging | ||
| [40] | DT, AdaBoost, LR, SGD, RF, SVM, GBM, ETC, G-NB | DT, AdaBoost, RF, GBM | Clinical | Oversampling | UCI Heart Failure | 13/299 | Bagging (DT) = 87.7%, Boosting (AdaBoost) = 88.5%, Bagging (RF) = 91.8%, Boosting (GBM) = 88.5% | Bagging | |
| [20] | LR, SVM, KNN, DT, RF | Majority Voting, RF, DT | Clinical | Handled missing values, imputation, normalisation | 139/164 | UCI Cleveland Heart Disease | 14/303 | Voting = 98.18%, Bagging (DT) = 93.1%, Bagging (RF) = 94.4% | Voting |
| [41] | NB, KNN, RT, SVM, BN | AdaBoost, LogitBoost, RF | Clinical | UCI SPECT heart disease | 22 | Bagging (RF) = 90%, Boosting (AdaBoost) = 85%, Boosting (LogitBoost) = 93% | Boosting | ||
| [7] | NB, RF, MLP, BN, C4.5, PART | Bagging, Boosting, and Stacking, | Clinical | Handled missing values | 139/164 | UCI Cleveland Heart Disease | 14/303 | Bagging = 79.87%, Boosting = 75.9% Stacking = 80.21%, Voting = 85.48% |
Voting |
| [42] | KNN, SVM, NB, LR, QDA, C4.5, NN | Bagging, AdaBoost, and Stacking | Clinical | 139/164 | UCI Cleveland Heart Disease | 14/303 | Bagging = 77.9%, Boosting (AdaBoost) = 64.3%, Stacking = 82.5% | Stacking | |
| [5] | LR, KNN, SVM, DT, NB, MLP | Bagging, Boosting, and Stacking | Equal | Kaggle Cardiovascular Disease | 12/- | Bagging = 74.42%, Boosting = 73.4%, Stacking = 75.1% | Stacking | ||
| [43] | RF, ET, XGBoost, GB | AdaBoost, GBM, Stacking | Eliminated outliers | IEEE Data Port | 11/1190 | Boosting (GBM) = 84.2%, Boosting (AdaBoost) = 83.4%, Stacking = 92.3% | Stacking | ||
| [44] | MLP, SCRL, SVM | Bagging, Boosting, and Stacking | Clinical | 139/164 | UCI Cleveland Heart Disease | 14/303 | Bagging = 80.5%, Boosting = 81.1%, Stacking = 84.1% | Stacking | |
| [28] | DT, CNN, NB, ANN, SVM, CAFL | Bagging, Boosting | Data distribution | Eric | 7/210 | Bagging = 73.2%, Boosting (AdaBoost) = 65.1%, Stacking = 79.4% | Stacking |