International Journal of Scientific and Research Publications, Volume 2, Issue 11, November 2012 440

ISSN 2250-3153

Fuzzy C- Means Algorithm- A Review

R.Suganya, R.Shanthi

Department of CS, Dr.SNS.Rajalakshmi College of Arts & Science

Abstract- Clustering is a task of assigning a set of objects into Fuzzy C-Means algorithm 2) The Gustafson-Kessel algorithm 3)
groups called clusters. In general the clustering algorithms can be The Gath-Geva algorithm. Shape based fuzzy clustering
classified into two categories. One is hard clustering; another one algorithm can be divided into 1) Circular shape based clustering
is soft (fuzzy) clustering. Hard clustering, the data’s are divided algorithm 2) Elliptical shape based clustering algorithm 3)
into distinct clusters, where each data element belongs to exactly Generic shape based clustering algorithm. In this paper, represent
one cluster. In soft clustering, data elements belong to more than a review on fuzzy c means, and extended version of fcm such as
one cluster, and associated with each element is a set of pcm, fpcm and their advantages and disadvantages of real time
membership levels. In this paper we represent a survey on fuzzy applications.
c means clustering algorithm. These algorithms have recently
been shown to produce good results in a wide variety of real
world applications. II. FUZZY C MEANS ALGORITHM
Fuzzy clustering is a powerful unsupervised method for
Index Terms- Soft clustering, hard clustering, FCM. the analysis of data and construction of models. In many
situations, fuzzy clustering is more natural than hard clustering.
Objects on the boundaries between several classes are not forced
I. INTRODUCTION to fully belong to one of the classes, but rather are assigned

F ast and robust clustering algorithms play an important role in

extracting useful information in large databases. The aim of
cluster analysis is to partition a set of N object into C clusters
membership degrees between 0 and 1 indicating their partial
membership. Fuzzy c-means algorithm is most widely used.
Fuzzy c-means clustering was first reported in the literature for a
such that objects within cluster should be similar to each other special case (m=2) by Joe Dunn in 1974. The general case (for
and objects in different clusters are should be dissimilar with any m greater than 1) was developed by Jim Bezdek in his PhD
each other[1]. Clustering can be used to quantize the available thesis at Cornell University in 1973. It can be improved
data, to extract a set of cluster prototypes for the compact by Bezdek in 1981. The FCM employs fuzzy partitioning such
representation of the dataset, into homogeneous subsets. that a data point can belong to all groups with different
Clustering is a mathematical tool that attempts to discover membership grades between 0 and 1.
structures or certain patterns in a dataset, where the objects inside
each cluster show a certain degree of similarity. It can be Algorithm
achieved by various algorithms that differ significantly in their
notion of what constitutes a cluster and how to efficiently find 1. Initialize U=[uij] matrix, U(0)
them. Cluster analysis is not an automatic task, but an iterative 2. At k-step: calculate the centers vectors C(k)=[cj] with
process of knowledge discovery or interactive multi-objective U(k)
optimization. It will often necessary to modify preprocessing and
parameter until the result achieves the desired properties.
In Clustering, one of the most widely used algorithms is
fuzzy clustering algorithms. Fuzzy set theory was first proposed 3. UpdateU(k) ,U(k+1)
by Zadeh in 1965 & it gave an idea of uncertainty of belonging 4.
which was described by a membership function. The use of fuzzy
dij
set provides imprecise class membership function. Applications
of fuzzy set theory in cluster analysis were early proposed in the
work of Bellman, Zadeh, and Ruspini This paper opens door step
of fuzzy clustering [2]. Integration of fuzzy logic with data
mining techniques has become one of the key constituents of soft 5. STOP; otherwise
computing in handling challenges posed by massive collections return to step 2.
of natural data. The central idea in fuzzy clustering is the non- 6.
unique partitioning of the data into a collection of clusters. The Here m is any real number greater than 1,
data points are assigned membership values for each of the
clusters and fuzzy clustering algorithm allow the clusters to grow uij is the degree of membership of xi in the cluster j,
into their natural shapes [3]. The fuzzy clustering algorithms can xi is the ith of d-dimensional measured data,
be divided into two types 1) Classical fuzzy clustering algorithms cj is the d-dimension center of the cluster,
2) Shape based fuzzy clustering algorithms. Classical fuzzy
clustering algorithms can be divided into three types.1) The

www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 2, Issue 11, November 2012 441
ISSN 2250-3153

This algorithm works by assigning membership to each Disadvantages

data point corresponding to each cluster center on the basis of 1) Very sensitive to good initialization
distance between the cluster center and the data point. More the 2) Coincident clusters may result
data is near to the cluster center more is its membership towards Because the columns and rows of the typicality matrix are
the particular cluster center. Clearly, summation of membership independent of each other
of each data point should be equal to one. After each iteration Sometimes this could be advantageous (start with a large
membership and cluster centers are updated according to the value of c and get less distinct clusters)
formula.

Advantages IV. FUZZY POSSIBILISTIC C MEANS ALGORITHM (FPCM)

1) Unsupervised To overcome difficulties of the pcm, Pal defines a
2) Converges clustering technique that integrates the features of both Fuzzy a
Possibilistic c-means called Fuzzy Possibilistic c-Means
Limitations: (FPCM).
1) Long computational time Membership and Typicality’s are very significant for the
2) Sensitivity to the initial guess (speed, local minima) accurate characteristic of data substructure in clustering
3) Sensitivity to noise and One expects low (or even no) difficulty. An objective function in the fpcm depending on both
membership degree for outliers (noisy points). membership and typicality’s are represented as::

Memberships and topicalities is represented as:

III. POSSIBILISTIC C-MEANS (PCM)
To overcome difficulties of the fcm, Krishnapuram and
keller proposed a new clustering model named Possibilistic c-
Means (PCM). Which of the following constraints

Algorithm

Fix the number of clusters C; fix m,1<m<∞;\

Set iteration counter l=1;
Intialize the possiblistic C-parttion U(0);
Estimate ηi FPCM generates Memberships and possibilities at the same time,
together with the usual point prototypes or cluster center for each
Repeat cluster.
Update the prototypes using U(l), as indicated bel ow;
Compute U(l+1) Advantage
Increment l; 1) Ignores the noise sensitivity deficiency of FCM
Until (|| U(l-1)-U(l)||<ε); 2) Overcomes the coincident clusters problem of PCM.

{ The remaining part of algorithm is optional and to be used only Disadvantages

when the actul shape of the generated possibility distribution is 1) The row sum constraints must be equal to one
important }

Set iteration counter l=1; V. POSSIBILISTIC FUZZY C MEANS ALGORITHM (PFCM)

Reestimate ηi In fpcm, the constraint corresponding to the sum of all
Repeat protypes using U(l), as indicated below; typicality values of all data to a cluster must be equal to one
Compute U(l+1) cause problems particularly for a big data set. In order to avoid
Increment l; this problem pal et al propose a new algorithm called
Until (|| U(l-1)-U(l)||<ε); Possibilistic Fuzzy c means algorithm (pfcm). The objective
function is defined by
ηi –determines distance at which the membership value of a point
in a cluster becomes 0.5.

Subject to

Advantage
Clustering noisy data samples , a&b define the relative importance between the membership
degrees and typicality values. The objective function can be

www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 2, Issue 11, November 2012 442
ISSN 2250-3153

minimized if for every i and k,m, >1

as well as z contains a minimum of c different data with these
AUTHORS
conditions we have (U,TT,V) .The membership
degree calculated with First Author – R.Suganya M.Sc., M.Phil., Assistant Professor,
Dr.SNS.Rajalakshmi College of Arts & Science,
Chinnavedampatti, Coimbatore., Email-
id:[email protected]
Second Author – R.Shanthi, Research Scholar, Dr. SNS.
Rajalakshmi College of Arts & Science, Chinnavedampatti,
Coimbatore. , Email-id:[email protected]
=1/1+( , 1<i<c

Advantage:
1. Ignores the noise sensitivity deficiency of FCM
2. Overcomes the coincident clusters problem of PCM.
3. Eliminates the row sum constraints of FPCM

VI. CONCLUSION AND FUTURE WORK

FCM algorithm is a distinctive clustering algorithm, has
been exploited in extensive range of engineering and scientific
disciplines, for instance, medicine imaging, pattern detection,
data mining and bioinformatics. In view of the fact, the initially
developed FCM makes use of the squared-norm to determine the
similarity between prototypes and data points, and it performs
well only in the case of clustering spherical clusters.
Furthermore, several algorithms are developed by numerous
authors based on the FCM with the aim of clustering more
general dataset. During the survey, we also find some points that
can be further improvement in the future using advanced
clustering technique to achieve more efficient accuracy in the
result and reduce the time taken for data and/or information
retrieval from large dataset.

REFERENCES
[1] M.S.Yang,” A Survey of fuzzy clustering” Mathl. Comput. Modelling Vol.
18, No. 11, pp. 1-16, 1993.
[2] A. vathy-Fogarassy, B.Feil, J.Abonyi”Minimal Spanning Tree based Fuzzy
clustering” Proceedings of World academy of Sc., Eng & Technology, vol-
8, Oct-2005, 7-12.
[3] Pal N.R, Pal K, Keller J.M. and Bezdek J.C, “A Possibilistic Fuzzy c-Means
Clustering Algorithm”, IEEE Transactions on Fuzzy Systems, Vol. 13, No.
4, Pp. 517–530, 2005.
[4] R. Krishnapuram amd J.M. Keller, “A possibilistic approach to clustering”,
IEEE Trans. Fuzzy Systems, Vol. 1, Pp. 98-110, 1993.
[5] Vuda Sreenivasarao and Dr.S. Vidyavathi, “Comparative Analysis of Fuzzy
C- Mean and Modified Fuzzy Possibilistic C -Mean Algorithms in Data
Mining”, IJCST Vol. 1, No. 1, Pp. 104-106, 2010.
[6] J. C. Dunn (1973): "A Fuzzy Relative of the ISODATA Process and Its Use
in Detecting Compact Well-Separated Clusters", Journal of Cybernetics 3:
32-57
[7] J. C. Bezdek (1981): "Pattern Recognition with Fuzzy Objective Function
Algorithms", Plenum Press, New York.
[8] Mohamed Fadhel Saad and Adel M. Alimi,” Modified Fuzzy Possibilistic
C-means,” Proceedings of the International MultiConference of Engineers
and Computer Scientists 2009 Vol I IMECS 2009, March 18 -20, 2009,
Hong Kong.

www.ijsrp.org