A Global-Relationship Dissimilarity Measure for the k-Modes Clustering Algorithm

Pseudocode 1

Pseudocodes of KBGRD algorithm.

Input: data set U and initial cluster number ;

Output: clusters.

Sub function Cluster(U, modes)

Begin:

(1) for   = to is the number of clusters.

(2)   for   = to is the number of objects.

(3)  Calculating according to Eqs. (2)–(6);

(4)    end for

(5) end for

(6) if ()

(7)   Classify th object into into th cluster;

End

Sub function Fun()

Begin:

(1) for   = to is the number of clusters.

(2)     for   = to is the number of objects.

(3)    Calculating SumDissimilarity according to Eq. (7);

(4) return SumDissimilarity;

end

Main function

Begin:

(1) Randomly choose distinct objects as initial modes from ;

(2) Cluster(, modes);

(3) newDissimilarity = Fun(); calculating the value of .

(4) Do

(5)  oldDissimilarity = newDissimilarity;

(6)  Update modes according to Eq. (9);

(7)  Cluster(, modes);

(8)  newDissimilarity = Fun();

(9) while newDissimilarity != oldDissimilarity;

End