Table 4.

Summary of reviewed techniques in Liver Disease.

Ref.	Base Learner Models	Ensemble Model	Data Type	Preprocessing Technique	Positive/Negative Cases	Dataset	Attributes/Instances	Accuracy	Best Model
[3]	BeggRep, BeggJ48, AdaBoost, LogitBoost, RF	Bagging, Boosting	Clinical		416/167	UCI Indian Liver Patient	10/583	Boosting(AdaBoost) = 70.2%, Boosting(LogitBoost) = 70.53%, Bagging (RF) = 69.2%	Boosting
[49]	NB, SVM, KNN, LR, DT, MLP	Stacking, DT	Clinical	Feature Selection PCA	416/167	UCI Indian Liver Patient	10/583	Bagging (DT) = 69.40% Stacking = 71.18%	Stacking
[50]	KNN	RF, Gradient Boosting, AdaBoost, Stacking	Clinical		416/167	UCI Indian Liver Patient	10/583	Bagging (RF) = 96.5%, Boosting(Gradient) = 91%, Boosting(AdaBoost) = 94%, Stacking = 97%	Stacking
[33]	DT, NB, KNN, LR, SVM, AdaBoost, CatBoost	XGBoost, Light GBM, RF	Clinical	Handled missing values	416/167	UCI Indian Liver Patient	10/583	Bagging (RF) = 88.5% Boosting(XGBoost) = 86.7% Boosting (LightGBM) = 84.3%	Bagging
[32]	SVM, KNN, NN, LR, CART, ANN, PCA, LDA	Bagging, Stacking	Clinical	Handled missing values, feature selection, PCA	453/426	Iris And Physiological	22/879	Bagging (RF) = 85%, Stacking = 98%	Stacking
[9]	KNN, SVM, RF, LR, CNN	RF, XGBoost, Gradient Boost		Handled missing values, scaling, and feature selection		Image	11/10,000	Bagging (RF) = 83% Boosting (XGBoost) = 82% Boosting (Gradient) = 85%	Boosting
[51]	LR, DT, RF KNN, MLP	AdaBoost, XGBoost, Stacking	Clinical	Data Imputation, label encoding, resampling, eliminating duplicate values and outliers	416/167	UCI Indian Liver Patient	10/583	Boosting (AdaBoost) = 83% Boosting (XGBoost) = 86% Stacking = 85%	Boosting
[10]	DT, KNN, SVM, NB	Bagging, Boosting, RF	Clinical	Discretisation, resampling, PCA	416/167	UCI Indian Liver Patient	10/583	Bagging (RF) = 88.6%, Bagging = 89%, Boosting = 89%	Bagging Boosting