This document discusses techniques to improve the efficiency of the K-means clustering algorithm. It begins with an introduction to K-means clustering and discusses some of its limitations, such as high computational time. It then proposes using a ranking method to help assign data points to clusters more efficiently. The key steps of standard K-means clustering and the proposed ranking-based approach are described. Experimental results on sample datasets show that the ranking method leads to faster clustering compared to standard K-means, with comparable accuracy. Therefore, the ranking method can help enhance the performance of K-means clustering for large datasets.

1. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 04 Issue: 10 | Oct -2017 www.irjet.net p-ISSN: 2395-0072
© 2017, IRJET | Impact Factor value: 6.171 | ISO 9001:2008 Certified Journal | Page 1438
A Study of Efficiency Improvements Technique for K-Means Algorithm
Rakesh P. Badgujar1, Mr. Shyam Sundar Meena2, Prof. Preetesh Purohit3
1Research Scholar, Computer Science & Engineering Department, Swami Vivekanand College of Engineering
Indore, India
2Asst. Professor, Computer Science & Engineering Department, Swami Vivekanand College of Engineering
Indore, India
3Professor, Head of Computer Science and Engineering Department, Swami Vivekananda College of Engineering,
Indore, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - K-Mean clusteringisthemethod whosedata set
divided K-number over cluster. The Determining outcomes
regarding their overall performance then K-meanclustering
in case over execution time discussed here. Butcertainsome
boundaries between K-mean clustering algorithm to take
argument time is more. So how many clustering to executed
less execution period so compared as like standard method,
also less the origin time for using the Ranking Method. This
approach accomplish an attempt because of analyzing the
feasibility concerning K-mean clustering algorithmforusing
Ranking Method between data mining.
Key Words: Round Robin, Consumption Laxity Based
1. INTRODUCTION
In today’s extraordinarily aggressive business environment
Clustering move an essential role.AsK-meansClusteringisa
method because of erection groups concerning the facts set
yet the objects to that amount are comparable properties.
2. CLUSTERING
Mainly Clustering is the approach who consists of the group
on similar type objects of certain lot or a brush as consists of
the objects of information employ is elected of rule in
conformity with limit partial dimension concerning
dissimilarity. Clusteringisa typeofunsupervisedlessonnow
not supervised instruction as Classification. In clustering
method, objects over the dataset are grouped within
clusters, between certain an access so much companies are
at all specific beside each other and the objects in the equal
team or bunch are at all comparable according toeachother.
o Hierarchical Clustering Algorithm
o K-means Clustering Algorithm
o Density Based Clustering Algorithm
o Self-organization maps
o EM clustering Algorithm
3. RELATED WORK
Below the table shows our Survey of reference papers for
using k-means algorithm.
Research
Paper
Improving
the Accuracy
and
Efficiency of
the k-means
Clustering
Algorithm
An Iterative
Improved k-
means
Clustering
Refining
Initial
Points for K-
Means
Clustering
Compariso
n of
various
clustering
algorithms
Problem
being
addressed
Lower
accuracy and
efficiency
Number of
Iterations
are Less
Estimate is
fairly
unstable
due to
elements of
the tails
appearing in
the sample
Which
clustering
algorithm
is best
Importanc
e of the
problem
algorithm
requires a
time
complexity
Total
number of
iterations
required by
k-means and
improved k-
means is
much larger
Importance
of the
problem of
having a
good initial
points
Way of
Process
Gap in the
prior
work
Accuracy and
Efficiency is
most
complicated
to reducing
Check
multiple
iterations
To finding
Initial
Points
Finding
algorithm
Specific
research
questions
or
research
objective
To Overcome
the problem
of Accuracy
and
Efficiency
This paper
presented
iterative
improved k-
means
clustering
algorithm
that makes
the k-means
more
efficient and
produce
good quality
clusters
A fast and
efficient
algorithm
for refining
an initial
starting
point for a
general
class of
clustering
algorithms
has been
presented
data
mining is
that to
discover
the data
and
patterns
and store
it in an
understan
dable
form
Broad
outline of
how the
author
solved the
problem
Using K-
Means
clustering
Algorithm
and The
enhanced
Method
Iteration
improve k-
means
cluster
algorithm
Using
Clustering
Cluster
Applied
DBSCAN
and
OPTICS
algorithms
Key
contributi
on of the
paper
claimed
by the
author.
define k
centroids,
one for each
cluster
iterative
improved k-
means
clustering
algorithm
Clustering
Clusters
K-Means
clustering
Algorithm

2. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 04 Issue: 10 | Oct -2017 www.irjet.net p-ISSN: 2395-0072
© 2017, IRJET | Impact Factor value: 6.171 | ISO 9001:2008 Certified Journal | Page 1439
4. K-Means Clustering Algorithm
K-means clustering is a ordinary partitioning method. In it
objects are categorized as belonginginimitationofoneon K-
groups. The consequences on partitioning method is a set
concerning K clusters, each goal of facts set belonging to one
cluster. In every tussock like may be a centroid or a tussock
representative. In suit the place we consider real-valued
data, the arithmetic vile regardingthevirtuevectors because
whole objects inside a tussock offers an suitable
representative; choice types regarding centroid might also
stay required of ignoble cases. Example: A lot on documents
be able be represented with the aid of a list over those
keywords as occur of half minimum variety of documents
inside a cluster. If the number on the clusters is large, the
centroids do lie further clustered in accordance with
produces hierarchy inside a dataset. K-means is a facts
excavation algorithm who performs clustering of the
information samples. As stated previously, clustering
capacity the division of a dataset in a range of agencies such
so much similar items fall yet belong according to identical
groups. In discipline in accordance with tussock the
database, K-means algorithm use an iterative approach. The
input of this litigation is the range regardingdesiredclusters
and the preliminary ability thenalsoproducesfinal potential
namely output. These observedinitial yetultimatecapability
are the ability about clusters. If ofthealgorithm requirement
is according to birth K clusters below even will stay K
preliminary ability then remaining capability also.
4.1 Measurement of Distance between Objects
and Means
In kilter to measurement the association within objectsthen
potential different K-means clustering strategies perform
remain used. Most popular without metric that aged is
Euclidean Distance. Euclideanscaleisrepresentedaslikethe
square bottom of addition concerning squared differences
among same degree of aim then the mean or cluster
centroid. Euclidean distance is the close common scale
metric who is close commonly aged now attitude including
multi-dimensional data.
4.2 Selection of Initial Means
Basically, the selection regarding initial ability is upontothe
developer regarding clusteringlawwherehe/shewants. But
that choice regarding initial capacity is unbiased concerning
K-means clustering, due to the fact this preliminary ability
are inputs about K-means algorithm. In some cases, such is
favored according to choose initial skill randomly out of the
given dataset while incomplete others prefer in accordance
with produce initial factors randomly. As regarded up to
expectation choice concerning preliminary potential affects
both the proof period regarding the algorithm or
additionally the godsend concerning K-means algorithm.
Certain techniques are added according to acquire better
effects up to expectation are considering the initial means.
a) The simplest shape concerning it techniques is that,
within order in imitation of accomplish K-means algorithm
along special sets about initial meansregardedyetsincepick
the auspicious results. But that method is rarely viable then
dataset is great then mainly because continuous K-means.
b) Another strategy to that amount is old in conformity with
acquire higher clustering outcomes is to utilizes refine
preliminary points method. If in case, that is possible to
begin K-means together with preliminary potential which
are closer to last means, below that is sharply viable action
so the range concerning iterations so the clustering
algorithm needs to converge intention decrease. It
additionally lessens the epoch required because
transformation then additionally will increase the precision
concerning closing means.
5. Steps of K-Means Clustering Algorithm
K-Means Clustering algorithm is an idea, among who at that
place is necessity according to align the given data put in
among K clusters, the cost of K (Number of clusters) is
defined via the person which is fixed. In this first the
centroid on each cluster is elected for clustering and
afterwards in accordance according to the select centroid,
the records points having minimum distanceoutofthegiven
cluster, is assigned to that uniquecluster.EuclideanDistance
is used because of calculating the scale regarding records
point from the precise centroid. This algorithm consists
about four steps:
I. Initialization
In this advance quadrant statistics set, wide variety
regarding clusters and the centroid so much we defined
because of each cluster.
II. Classification
The strip is deliberated because every facts point beside the
centroid then the facts factor abject minimum range beside
the centroid on a cluster is assigned in accordance with as
specific cluster.
III. Centroid Recalculation
Clusters generated previously, the centroid is again oft
thought capacity recalculation regarding the centroid.
IV. Convergence Condition
Some convergence prerequisites are attached as below:
a) Stopping then reaching a fond yet defined variety over
iterations.
b) Stopping so at that place is no trade regarding statistics
factors within the clusters.
c) Stopping so a introduction price is achieved.

3. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 04 Issue: 10 | Oct -2017 www.irjet.net p-ISSN: 2395-0072
© 2017, IRJET | Impact Factor value: 6.171 | ISO 9001:2008 Certified Journal | Page 1440
V. If whole about the on prerequisites arenolongersatisfied,
after go according to step 2 yet the total technique repeat
again, until the partial stipulations are no longer satisfied.
6. Ranking Method
With regards according to Clustering,rankingoperations are
a natural way to estimatethe probabilityovertheprevalence
regarding statistics items then the objects. So we
recommend evaluating ranking ordinarysketchondatabase
because of scholar informationbetweendisciplineaccording
to form the clusters. So Ranking feature introduce instant
opportunities in imitation of optimize the effects about K-
means clustering algorithm.
6.1 Need about Ranking Method
Search regarding applicable files and comparable records
inquire is a most popular feature over database in
conformity with attain knowledge. There are definitive
similar archives so much we want to read of some class then
shape certain cluster. That`s why, we want in imitation of
rank the greater relevance pupil marksbya ratingtechnique
then according to enhance inquire effectiveness. In last,
related solutions intention remain again because of a given
key-word question by means of the built index and higher
rating strategy. So I hold applied that Ranking approach
together with K-means clustering approach due to the fact
that approach is also abject the faith in accordance with
locate relevant records. So such is additionally useful into
creating clusters so much are base comparableresidences in
every data factors within to that amount cluster.
7. TOOLS USED FOR K-MEANS CLUSTERING
ALGORITHM IMPLEMENTATION
The equipment that are used because the implementation
concerning this expanded k-means clustering algorithm
included with beginning value and additionally because
Ranking Method is the Visual Studio 2008 the usage of C#.
8. RESULTS
8.1. K-Means Clustering Results
In it case, clusters are tooled into K-means clustering
algorithm, the use of the thoughtoverthresholdvalue.Graph
so is addicted beneath indicates the range over clusters as
are done about the foundation concerning the threshold
value. On the groundwork over the centroid the clusters are
formed. This layout is made concerning the groundwork on
the values x yet y, who values are made regarding the each
bead over the graph. The Euclidean strip is considered into
both the centroid then the statistics points. Each cluster is
shown along specific color of method in conformity with
discriminate between them.
8.2 K-Means Clustering Results using Ranking
Method
Graph below suggests the propriety then overall
performance on ranking method. In that case, clusters are
built of the groundwork regarding office so is considered by
way of making use of rating method. The solution epoch
additionally reduces as compared in conformity with K-
means clustering algorithm yet that is aged on significant
facts set. As shown among graph, the clusters are
constructed with rigor then properly differentiated out of
each- other.
8.3 Execution Time Analysis For K- Means
Clustering Algorithm
Execution era analysis for K-means clustering algorithm is
instituted of the foundation regarding the variety of files to
that amount are considered for clustering or or a good deal
day is done by way of this complete process.
In the table so much also indicates the number on data yet
the clustering knowledge era instituted through K-means
clustering algorithm is shown. As proviso the number of
archives are 50, the solution day wish lie 98ms then
consequently on. With the assist regarding this kind about
tables we may easily tell the performance.
Table -1: Execution time for K-means clustering
Records
Execution Time for Clustering
Method
50 98
100 132
150 198
200 209
250 287
300 309
350 380
400 390
450 467
500 487
8.4. Execution Time Analysis for using Ranking
Method
The argument time for ranking method is less. So this is an
excellent strategy utilized because clustering method. As in
action concerning solely K-means clustering because 50
information smoke the knowledge period up to expectation
is 98ms, however within that action concerning Ranking
method, for the cause on executive equal variety of records,
that takes 91ms. And the major desk so much shows the
origin time because the Ranking method because of each
precise archives

4. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 04 Issue: 10 | Oct -2017 www.irjet.net p-ISSN: 2395-0072
© 2017, IRJET | Impact Factor value: 6.171 | ISO 9001:2008 Certified Journal | Page 1441
Table -2: Execution time table for ranking method
Records
Execution time for Ranking
Method
50 91
100 121
150 167
200 190
250 267
300 310
350 326
400 376
450 422
500 476
9. CONCLUSIONS
The proposed job represents ranking primarily based
technique as extended K-means clustering algorithmoverall
performance yet accuracy. In this we hold additionally
committed analysis over K-means clustering algorithm
including the aid about applying couple methods, certain is
the current K-means clustering strategy which is
incorporated along some city worth afterward 2d some is
rating approach applied on K-means algorithm below
additionally in contrast the performance on every the
strategies by the utilization concerning graphs. The
experimental results validated so the proposed ranking
based K-means algorithmproduceshighereffectsthanhence
a whole lot on the current k-means algorithm.
10. Future Work
In future, within action about clustering the marks over
students beyond different-2 databases are considered by
using the use of the notion regarding Query redirection. By
the use of the Query redirection approach we execute easily
fascicle the sizeable total about facts from dispensed
environment so beyond distinct databases. So proviso this
approach is considered, afterwards the performanceoverK-
means clustering algorithm is accelerated because
substantial samples concerning records put in to that
amount are also allotted in nature.
ACKNOWLEDGEMENT
Foremost, I would like in accordance with express my
sincere thanks in imitation of Mr. S. S. Meena sir whichever
gave him courage whelming complete help among the
ending concerning that lookup delivery note including him
stimulating suggestions or development in conformity with
suffice beforehand into whole the time.Hehasalwaysbeena
supply of notion yet confidence because of me. He has
beaconed mild in imitation of me as much a guide at every
degrees regarding training concerning my Research work. I
categorical mythanksgiving besidetheinterior regarding my
morale in accordance with my dad and mom yet pals
because of encouragement, assistance or additionally help
between challenging circumstances. At last I am at all
thankful in imitation of my GOD anybody has given me it
white chance in conformity with operate M.Tech namely
nicely as like in imitation of slave research work.
REFERENCES
[1] K. A. Abdul Nazeer & M. P. Sebastian” Improving the
Accuracy and Efficiency of the K-Means Clustering
Algorithm” .Proceedings of the World Congress on
Engineering 2009 Vol I WCE 2009, London, U.K, July 1 -
3, 2009.
[2] D. Napoleon & P. Ganga lakshmi, “An Efficient K-Means
Clustering Algorithm for Reducing Time Complexity
using Uniform Distribution Data Points”, IEEE, 2010.
[3] Madhuri A. Dalal & Nareshkumar D. Harale “AnIterative
Improved k-means Clustering” Proc. of Int. Conf. on
Advances in Computer Engineering, 2011.
[4] Paul S. Bradley & Usama M. Fayyad, “Refining Initial
Points for K-Means Clustering”, 15th International
Conference on Machine Learning, ICML98.
[5] Osama Abu Abbas “Comparison of various clustering
algorithms” The International Arab Journal of
Information Technology, Vol. 5, No. 3, July 2008.
[6] Jirong Gu & et.al, “An Enhancement of K-means
Clustering Algorithm “, IEEE International Conference
on Business Intelligence and Financial Engineering,
2009.
[7] Dost Muhammad Khan & Nawaz Mohamudally “A
Multiagent System (MAS) for the Generation of Initial
Centroids for k-means clusteringData MiningAlgorithm
Based on Actual Sample datapoints”, IEEE, 2009.
[8] Malay K. Pakhira, “Clustering Large Databases in
Distributed Environment “, IEEE 2009 WEE
International.
[9] Shi Na & et.al,“Research on k-means Clustering
Algorithm”, IEEE Third International Symposium on
Intelligent Information Technology and Security
Informatics, 2010.