Introduction to Pattern Recognition: A Matlab Approach

124 CHAPTER 4 Feature Selection

Step 2. To rank the features use their cross-correlation with the compositeFeaturesRanking function:

featureNames = {'mean ','st. dev.','skewness','kurtosis'};

a1=0.2;a2=0.8;

[p]= compositeFeaturesRanking (class1,class2,a1,a2,T);

Print out the results, t ypi ng

fprintf('\n Scalar Feature Ranking \n');

for i=1:size(T,1)

fprintf('(%10s) \n',featureNames{T(i,2)});

end

fprintf('\n Scalar Feature Ranking with correlation \n');

for i=1:size(p,1)

fprintf('(%10s) \n',featureNames{p(i)});

end

The program output is

Scalar Feature Ranking

( mean )

( st. dev.)

( skewness)

( kurtosis)

Scalar Feature Ranking with correlation

( mean)

( kurtosis) ...