Effective Feature Selection for 5G IM Applications Traffic Classification

Algorithm 1

Feature selection algorithm based on weighted mutual information combined with ACC (WMI_ACC).

          Input:     // training data set,

     Output:  feature               //selected feature set

     (1) begin

     (2) for   to  

     (3)    calculate weight value for each features;

     (4) end for

     (5) for   to  ;

     (6)    calculate ;

     (7)    if ;

     (8)     insert into descending order;

     (9)    end if

    (10)    end for

    (11)     = getfirstfeatures(list);

    (12)    end until ( == NULL);

    (13)     is a data set of samples

        Values of features;

    (14)   last_accuracy ← classify ;

    (15)   Insert the feature into ;

    (16)   Feature = get next features;

    (17)   For feature is not NULL

    (18)     insert the feature into ;

    (19)      is a data set of samples values for ;

    (20)     Accuracy ← classify with a specific classifiers;

    (21)     If (ACC <= last_ACC)

    (22)        Remove features from ;

    (23)     else

    (24)        feature = getNextfeature;

    (25)   end if

    (26)   end for

return  ;