Table 1.
Number of genes, samples, and class cardinality in each cancer-related microarray data set.
Fig 1.
Simplified example of a nested stratified fivefold cross-validation.
Table 2.
The average test set accuracy rates and the average numbers of selected genes of ten runs of fivefold cross-validation obtained by BIRS, BIRSR, IGIS, and IGIS+ using the KNN classifier as the classifier.
Fig 2.
The box plots of the average accuracy rates (for n = 10 samples) obtained by BIRS, BIRSR, IGIS, and IGIS+ on each microarray data set using the KNN classifier as the classifier.
The whiskers are extended to the most extreme data points that are not outliers.
Fig 3.
The bar chart of the average numbers of wrapper evaluations of ten runs of fivefold cross-validation required by BIRS, BIRSR, IGIS, and IGIS+ on each microarray data set using the KNN classifier as the classifier.
Table 3.
The average F-scores and the average numbers of selected genes of ten runs of fivefold cross-validation obtained by BIRS, BIRSR, IGIS, and IGIS+ using the KNN classifier.
Table 4.
The average test set accuracy rates and the average numbers of selected genes of ten runs of fivefold cross-validation obtained by BIRS, BIRSR, IGIS, and IGIS+ using the decision tree as the classifier.
Fig 4.
The box plots of the average accuracy rates (for n = 10 samples) obtained by BIRS, BIRSR, IGIS, and IGIS+ on ten microarray data sets using the decision tree as the classifier.
Fig 5.
The bar chart of the average numbers of wrapper evaluations of ten runs of fivefold cross-validation required by BIRS, BIRSR, IGIS, and IGIS+ on each microarray data set using the decision tree classifier.
Table 5.
The average F-scores and the average numbers of selected genes of ten runs of fivefold cross-validation obtained by BIRS, BIRSR, IGIS, and IGIS+ using the decision tree classifier.
Fig 6.
Venn diagrams of genes selected by the four gene selection algorithms using the KNN classifier on (A) the colon tumor data set and (B) the ALL-AML-3 data set.