Affinity propagation based clustering algorithm
will be individually placed on each and every object
Specific cluster. Utilizing the subsequent clustering
technique. Affinity Propagation (AP) clustering continues to
be proven to work in many of clustering problems. This
particular paper views the best way to utilize AP in
incremental clustering problems. Firstly, we mention the
problems within Incremental Affinity Propagation (IAP)
clustering, then propose two techniques to solve them.
Correspondingly, two IAP clustering algorithms are usually
proposed. Five popular labeled data sets, real-world time
series as well as a video are employed test the performance
associated with IAPKM and IAPNA. Standard AP
clustering is usually implemented to produce benchmark
performance. Experimental results show that IAPKM and
IAPNA is capable of doing comparable clustering
performance together with standard AP clustering on each
of the data sets. Meanwhile, the time cost is actually
dramatically reduced within IAPKM and IAPNA. The two
effectiveness and also the efficiency make IAPKM and
IAPNA capable of being well utilized in incremental
clustering tasks.
Published In : IJCAT Journal Volume 2, Issue 8
Date of Publication : August 2015
Pages : 288 - 292
Figures :01
Tables : 01
Publication Link :An Improved Dynamic Based Incremental
Clustering in Affinity Propagation
P. Sujitha : M.Tech 2nd year, Department of CSE, JNTUA, CREC
Tirupati, AP, India
R. Suresh : Professor & Head , Department of CSE, JNTUA, CREC
Tirupati, AP, India
Cluster
affinity propagation
IAP
IAPNA
IAPKM
This paper describes a new and novel approach for
incremental K-means clustering. The method stems from
the pyramid K-means algorithm presented in difference
from the pyramid approach, the sampling is done without
replacement. Furthermore, the sampling size is fixed. On
the other hand, two measures are applied to the data in
order to overcome the fact that each data block is
processed only one time. First, the algorithm starts with a
relatively large number of clusters and scales the number
down in U1e last stage. the dynamic approach can be used
to mitigate another inherent problem of K-means where
the number of clusters has to be predetermined. In both
cases the algorithms work on "chunks" of data referred to
as blocks.
[1] Xindong Wu, Vipin Kumar,J.Ross Quinlan, Joydeep
Ghosh, Qiang Yang, Hiroshi Motoda,Geoffrey, J.
McLachlan, Angus Ng, Bing Liu, Philip S. Yu, Zhi-Hua
Zhou, Michael steinbach,David J. Hand, and Dan
Steinberg, 2007. A survey paper on Top 10 algorithm in
data mining,published by Springer.
[2] M. Mehta, R. Agrawal, and J. Rissanen, 1996. SLIQ: A
fast scalable classifier for data mining.In Proceeding of
International Conference on Extending Database
Technology (EDBT?96), Avignon, France.
[3] H J. Shafer, R. Agrawal, and M. Mehta, 1996. SPRINT:
A scalable parallel classifier for data mining. In
Proceedings of the 22 nd International Conference Very
Large Databases (VLDB),pages 544-555, Mumbai, India.
New York: Springer-Verlag, 1985, ch. 4
[4] J. Han, and M. Kamber, 2006. Data Mining Concepts
and Techniques, Elsevier Publishers.
[5] H. Freeman, Jr., 1987. Applied Categorical Data
Analysis, Marcel Dekker, Inc., New York. Micheline
Kamber, Lara Wistone, Wan Gong, Shan Chen, and Han,
1997. Generalization and Decision Tree Induction:
Efficient classification in Data Mining, for IEEE.
