. 2022 Nov 2;37(3):217–226. doi: 10.4103/ijnm.ijnm_111_21

Table 1.

Attribute selection and machine-learning classification methods along with the protocol used in the study

Attribute evaluator + search method	Classifier
1. CFS subset evaluator + greedy stepwise 2. Information gain attribute evaluator + ranker 3. Principal components analysis + ranker* 4. ReliefF attribute evaluator + ranker - number of nearest neighbors (k): 10 Equal influence nearest neighbors 5. OneRules attribute evaluator + ranker 6. Mann–Whitney U-test with P<0.05 7. No selection	1. Naive Bayes updateable 2. OneRules 3. Multinomial logistic regression using a ridge estimator - logistic regression with ridge parameter of 1.0E-8 4. Simple logistic regression 5. Multilayer perceptron using sigmoid nodes 6. Logistic model tree – LM_1: 6/6 (40) Number of leaves: 1, size of the tree: 1 7. Random forest - bagging with 100 iterations and base learner 8. AdaBoostM1, weight: 0.25 Number of performed iterations: 10 9. Bagging – bagging with 10 iterations and base learner 10. Iterative classifier optimizer 11. Randomizable filtered classifier - IB1 instance-based classifier using 1 nearest neighbor (s) for classification 12. K-Nearest neighbors – IB1 instance-based classifier using 1 nearest neighbour (s) for classification 13. Support vector machine using sequential minimal optimization - IB1 instance-based classifier using 1 nearest neighbor (s) for classification 14. LogitBoost – number of performed iterations: 10

Attribute evaluator + search method

Classifier

1. CFS subset evaluator + greedy stepwise
2. Information gain attribute evaluator + ranker
3. Principal components analysis + ranker*
4. ReliefF attribute evaluator + ranker - number of nearest neighbors (k): 10
Equal influence nearest neighbors
5. OneRules attribute evaluator + ranker
6. Mann–Whitney U-test with P<0.05
7. No selection

1. Naive Bayes updateable
2. OneRules
3. Multinomial logistic regression using a ridge estimator - logistic regression with ridge parameter of 1.0E-8
4. Simple logistic regression
5. Multilayer perceptron using sigmoid nodes
6. Logistic model tree – LM_1: 6/6 (40)
Number of leaves: 1, size of the tree: 1
7. Random forest - bagging with 100 iterations and base learner
8. AdaBoostM1, weight: 0.25
Number of performed iterations: 10
9. Bagging – bagging with 10 iterations and base learner
10. Iterative classifier optimizer
11. Randomizable filtered classifier - IB1 instance-based classifier using 1 nearest neighbor (s) for classification
12. K-Nearest neighbors – IB1 instance-based classifier using 1 nearest neighbour (s) for classification
13. Support vector machine using sequential minimal optimization - IB1 instance-based classifier using 1 nearest neighbor (s) for classification
14. LogitBoost – number of performed iterations: 10

*All feature selection and classification methods were 10-fold cross-validated (stratified) with seed value=1 except for principal components analysis + ranker. Confusion matrices were obtained for each classifier along with performance metrics. CFS: Correlation-based feature selection