A Mixed Feature Selection Method Considering Interaction

Algorithm 1

NIWFS: neighborhood interaction weight based feature selection algorithm.

Input: Dataset with original feature set and the class

    Number of selected feature , neighborhood size

Output: Selected feature subset

(1) ;

(2) ;

(3) Initial the weight to 1 for each feature;

(4) For each do

(5)  Calculate using Definition 8;

(6) End

(7) While do

(8)  For each candidate feature do

(9)   Calculate the adjusted relevance measure ;

(10)  End

(11)  Select the feature with the largest ;

(12)   ;

(13)   ;

(14)  For each candidate feature do

(15)   Calculate the interaction weight factor ;

(16)   Update ;

(17)  End

(18)   ;

(19) End