A Novel Clustering Approach Using Hadoop Distributed Environment - Computational Intelligence Techniques for Comparative Genomics - page 110

Fig. 2 Procedure for canopy
clustering algorithm

distance measure used to create canopies are approximate, the obtained clusters are
not accurate. Hence, for obtaining much accurate clusters, the obtained output
canopies are fed as input to the fuzzy C-means algorithm. Earlier literature
addresses the other algorithms [ 8 ] like K-Means ef ciently performs clustering by

cient when the number of clusters is
large. Another disadvantage of K-Means is that it works effectively only for smaller
datasets.
Hence, in this paper, we have utilized the Soft-Fuzz (Fuzz C-Means) algorithm
for ef

nding good initial starting point, but is not ef

cient clustering. The algorithm for Fuzzy C-Means clustering is presented in
the next section.

4 Fuzzy C-Means Clustering Algorithm

The Fuzzy C-Means clustering applies fuzzy partitioning process so that the data
point might belong to all groups having different membership grades ranging from
0 to 1 and is iteratively performed. Finding cluster centers that minimize dissimi-
larity function is the main aim of Fuzzy C-Means algorithm. For fuzzy partitioning,
the membership matrix (U) is being randomly initialized as per Eq. ( 1 ).

X

c

u ij ¼ 1 ; 8 j ¼ 1 ; ... ; n ð 1 Þ

i ¼ 1

The dissimilarity function used is given as

X X X

c c n

u ij d ij J ð U ; c 1 ; c 2 ; ... ; c c Þ¼ J i ¼ ð 2 Þ

i ¼ 1 i ¼ 1 j ¼ 1

where 0 < u ij <1