In line 2, the parent cluster to be split is initialized. In lines 4–7, a 2-means clustering algorithm
is run I times to determine the best split which maximizes the Ward’s distance between C1 and C2.
In lines 9–10, the best split obtained will be used to divide the parent cluster. In line 11, the larger
of the split clusters is made the new parent for further splitting. The computational complexity of
the bisecting K-means is much higher compared to the standard K-means.