Table 3.

Summary of reviewed techniques in Skin Cancer. The positive/negative cases come in a range since this dataset contains data for six different skin cancer conditions.

Ref.	Base Learner Algorithm	Ensemble Approach	Data Type	Preprocessing Technique	Positive/Negative Cases	Dataset	Attributes/Instances	Accuracy	Best One
[18]	NB, RF KNN, SVM, and MLP	Bagging, Boosting, and Stacking	Clinical	Handled missing values	[20–112]/[254–346]	UCI Dermatology	34/366	Bagging = 96%, Boosting = 97%, Stacking = 100%	Stacking
[8]	DT, LR	Bagging, AdaBoost, and Stacking	Clinical	Feature selection	[20–112]/[254–346]	UCI Dermatology	34/366	Bagging = 92.8%, Boosting (AdaBoost) = 92.8%, Stacking = 92.8%	Bagging Boosting Stacking
[48]	LR, CHAID DT	Bagging, Boosting	Clinical	Handled missing values, data distribution, and balancing,	[20–112]/[254–346]	UCI Dermatology	34/366	Bagging = 100%, Boosting = 100%	Bagging Boosting
[31]	NB, KNN, DT, SVM, RF, MLP	Bagging, Boosting, and Stacking	Clinical	Hybrid Feature selection, information gain, and PCA	[20–112]/[254–346]	UCI Dermatology	12/366	Bagging = 95.94%, Boosting = 97.70%, Stacking = 99.67%	Stacking
[16]	PAC, LDA, RNC, BNB, NB, ETC	Bagging, AdaBoost, Gradient Boosting	Clinical	Feature Selection	[20–112]/[254–346]	UCI Dermatology	34/366	Bagging = 97.35%, AdaBoost = 98.21%, Gradient Boosting = 99.46%	Boosting