
4.8 Feature Subset Selection 129
2. Select the 14 highest-ranked feat ures and empl oy the exhaustive search method with t he J
3
criterion
to select the best combination of two features.
3. Repeat step 2 for SFS, SBS, and SFFS and comment on the results obtained.
Solution. To load the data sets of the two classes, type
class_1=load('testClass1.dat')';
class_2=load('testClass2.dat')';
and t hen do the following:
Step 1. Normalize and rank features in FDR descending order, t ypi ng
[class_1,class_2]=normalizeStd(class_1,class_2);
[TFisher]=ScalarFeatureSelectionRanking(...
class_1,class_2,'Fisher');
[pFisher]= compositeFeaturesRanking (...
class_1,class_2,0 ...