International Journal of Science and Research (IJSR)

ISSN (Online): 2319-7064

Index Copernicus Value (2015): 78.96 | Impact Factor (2015): 6.391

K- Means Clustering Algorithm Applications in

Data Mining and Pattern Recognition
Huda Hamdan Ali1, Lubna Emad Kadhum2
1, 2
Assistant Lecturer, computer Engineering Techniques,Al-Imam Alkadhum College, Baghdad-Iraq

Abstract: Clustering is a process of partitioning a set of data (or objects)into a set of meaningful sub-classes, called clusters, help users
understand the natural grouping or structure in adata set. Clustering has wide applications, inEconomic Science (especially market
research), Document classification,Pattern Recognition, Spatial Data Analysis and Image Processing.This paper focuses on clustering
in data mining and image processing. K-means algorithm is the chosen clustering algorithm to study in this work. The paper include:
the algorithm and its implementation, how to use it in data mining application and also in pattern recognition.

Keywords: k-means, clustering, data mining, pattern recognition

1. Introduction objects in other clusters. A cluster of dataobjects can be

treated collectively as one group and so may be considered
The k-means algorithm is the most popular clustering tool as a formof datacompression. Although classification is an
used in scientific and industrial applications[1]. The k-means effective means for distinguishing groups orclasses of
algorithm is best suited for data miningbecause of its objects, it requires the oftencostly collectionand labeling of
efficiency in processing large data sets. Clustering is one of a large set of trainingtuples or patterns,which the classifier
the well-known Data miningtechniques to find useful pattern uses tomodel each group.Clustering is also called data
from a data in a largedatabase (Fayyad, 1996)[2],However segmentation in some applications because clustering
working only on numeric values limits its use indata mining partitions large data sets into groups according to their
because data sets in data mining often havecategorical similarity.
values.
A good clustering method will produce high qualityclusters
The Data mining approaches has been known to aid the whichthe intra-class (that is, intra-cluster) similarity is
process of detecting an intrusion in network environment. high,the inter-class similarity is low,the quality of a
Thus, clustering algorithms were widely used for intrusion clustering result also depends on boththe similarity measure
detection. M. Jianlian[3] introduced the application on used by the method and itsimplementation,the quality of a
intrusion detection based on K- means clustering algorithm. clustering method is also measured byits ability to discover
K-means is used for intrusion detection to detect unknown some or all of the hidden patterns,however, objective
attack. Lei Li, et al [4] introduced a novel rule-based evaluation is problematic usuallydone by human / expert
intrusion detection system using data mining. They proposed inspection.
an improvement over priory algorithm by bringing the
concept of length-decreasing support to detect intrusion It's In general, the major clustering methods can be classified
also used in image processing generally and especially in into the following categories Partitioning methods,
pattern recognition applications programs to achieve some Hierarchical methods, Density-based methods, Grid-based
purpose like segmentation, where it gave an efficient methods, Model-based methods.
solution for this important case.Alan Jose, S. Ravi and
M.Sambath[5],proposed Brain Tumor Segmentation using 3. K-Means
K-means Clustering and FuzzyC-means Algorithm and its
area calculation. In the paper, they divide the process into The most well-known and commonly used partitioning
three parts, pre-processing ofthe image, advanced k-means methods are k-means.The k-means algorithm takes the input
and fuzzy c-means and lastly the feature extraction. First parameter, k, and partitions a set of n objects into k clusters
pre-processing is implementedby using the filter where it so that the resulting intra cluster similarity is high but the
improves the quality of the image. Then the proposed inter cluster similarity is low. Cluster similarity is measured
advance K-means algorithm is used, followed by Fuzzy c- in regard to the mean value of the objects in a cluster, which
means to cluster the image. Then the resulted segment image can be viewed as the cluster’s centroid or center of
is used for the feature extractionfor the region of interest. gravity.“How does the k-means algorithm work?” The k-
They used MRI image for the analysis and calculate the size means algorithm proceeds as follows.
of the extracted tumor region inthe image.
First, it randomly selects k of the objects, each of which
2. Clustering[1] initially represents a cluster mean or center. For each of the
remaining objects, an object is assigned to the cluster to
The process of grouping a set of physical or abstract objects which it is the most similar, based on the distance between
into classes of similar objectsis called clustering. A cluster the object and the cluster mean. It then computes the new
is a collection of data objects that are similar to one mean for each cluster. This process iterates until the criterion
anotherwithin the same cluster and are dissimilar to the function converges [1].K-Means clustering is one of the
Volume 6 Issue 8, August 2017
www.ijsr.net
Licensed Under Creative Commons Attribution CC BY
Paper ID: ART20176024 DOI: 10.21275/ART20176024 1577
 International Journal of Science and Research (IJSR)
ISSN (Online): 2319-7064
Index Copernicus Value (2015): 78.96 | Impact Factor (2015): 6.391
most famous clustering algorithms applied in different types
of domains such as Biology and Zoology, Medicine and
Psychiatry, Sociology and Criminology,
Geology,Geography and Remote sensing, Pattern
recognition and market research, and Education etc…[6].

3.1 Algorithm: k-means

The k-means algorithm for partitioning, where each

cluster’scenter is represented by the mean value of the
objects in the cluster.

Input:
k: the number of clusters,
D: a data set containing n objects.
Output: A set of k clusters.

Method:
1) Arbitrarily choose k objects from D as the initial cluster
centers; Figure 1: Flow chart of k-means algorithm
2) Repeat
3) (Re) assign each object to the cluster to which the object 3.2 k-mean example
is the most similar,based on the mean value of the objects
in the cluster; The below figures shows the steps of implementing k-means
4) Update the cluster means, i.e., calculate the mean value algorithm in details.
of the objects foreach cluster;
5) Until no change;

Volume 6 Issue 8, August 2017

www.ijsr.net
Licensed Under Creative Commons Attribution CC BY
Paper ID: ART20176024 DOI: 10.21275/ART20176024 1578
 International Journal of Science and Research (IJSR)
ISSN (Online): 2319-7064
Index Copernicus Value (2015): 78.96 | Impact Factor (2015): 6.391

4. K-Mean Algorithm and Data Mining uniquecomputational requirements on relevant clustering

algorithms. A variety ofalgorithms have recently emerged
The biggest advantage of the k-means algorithm in that meet these requirements and weresuccessfully applied
datamining applications is its efficiency in clustering to real-life data mining problems. [8]
largedata sets [7].Data mining adds to clustering the
complications of very largedatasets with very many Today’s business world is fast and dynamic in nature. It
attributes of different types. This imposes involves lot of data gathered from different sources.These
data are stored in Data Warehouses. The mostchallenging
Volume 6 Issue 8, August 2017
www.ijsr.net
Licensed Under Creative Commons Attribution CC BY
Paper ID: ART20176024 DOI: 10.21275/ART20176024 1579
 International Journal of Science and Research (IJSR)
ISSN (Online): 2319-7064
Index Copernicus Value (2015): 78.96 | Impact Factor (2015): 6.391
task of the business people is to transformthese data into real world data base “Machinelearning repository, 2012”.
useful information called knowledge. Datamining techniques One of data set they used was “Iris Dataset” (Fisher,1936). It
are used to achieve this task. So to illustrate this case some consists of 5 attributes and150 instances. Theattributes are
examples are discussed in this paper. sepal width, sepal length, petal width, petallength and class
label. It has three classes Iris flowers namely: C Iris setosa,
4.1using of K-Means Clustering in Cloud Computing C Iris versicolor,C Iris virginica. Theyhave created a
Environment[6] database called “KMean” and twotables to store the data sets
of iris and Blood TransfusionService Center Data Set in
Cloud Computing offers several benefits to the business Cloud SQL using MySQL. Asone of the properties of K-
organization to cut the initial investments to establish Mean is that it assumes thenumber of clusters K is known in
infrastructure for storage and compute. Many business advance.
organizations have already started migration of their
business data into cloud data centers. Most often they need For the first experiment, they assume that k = 3.Figure 2
to mine useful information from the data stored in the cloud shows the output clusters of Iris dataset obtainedfrom the
data centers with regards to business decisions. So the main experiment. Since the value of K is 3 thenumbers of output
objective of this work was to incorporate and implement K- clusters are three. The contents of eachcluster are displayed
Means Data mining technique into Cloud environment.For under the name cluster1, cluster2 andcluster3.
the experiment they took two data sets from the well-known

Figure 2: k-means action in cloud experiment on IRIS data set

So and after implementation of k- mean algorithm on that bioinformatics, weatherforecasting, fraud detection, financial
data set the result was:K-Means algorithm is more efficient analysis and customersegmentation. Customer segmentation
algorithmfor mining large Databases and Cloud computing is the subdivision of a businesscustomer base into groups
providessolution for storing largedatabase with less cost. called customer segments such thateach customer segment
consists of customers who sharesimilar market characteristics.
4.2 using K-Means Algorithm for EfficientCustomer This segmentation is based onfactors that can directly or
Segmentation[9] indirectly influence market orbusiness such as products
preferences or expectations,locations, behaviors and so on.
Data mining is the process of extracting meaningful The importance of customer segmentation include, inter alia,
information from dataset and presenting it in a human the ability of a business tocustomize market programs that
understandable format forthe purpose of decision support. will be suitable for each of itscustomer segments; business
The data mining techniquesintersect areas such as statistics, decision support in terms ofrisky situation such as credit
artificial intelligence,machine learning and database systems. relationship with its customers.
The applications ofdata mining include but not limited to

Volume 6 Issue 8, August 2017

www.ijsr.net
Licensed Under Creative Commons Attribution CC BY
Paper ID: ART20176024 DOI: 10.21275/ART20176024 1580
 International Journal of Science and Research (IJSR)
ISSN (Online): 2319-7064
Index Copernicus Value (2015): 78.96 | Impact Factor (2015): 6.391
Clustering has proven efficient in discovering subtle but
tactical patterns or relationships buried within a repository of
unlabeled datasets. This form of learning is classified under
unsupervised learning. Clustering algorithms include k-Means
algorithm, k-Nearest Neighbor algorithm, Self-Organizing
Map (SOM) and so on. These algorithms, without any
knowledge of the dataset beforehand, are capable of
identifying clusters therein by repeated comparisons of the
input patterns until the stable clusters in the training examples
are achieved based on the clustering criterion or criteria. Each
cluster contains data points that have very close similarities
but differ considerably from data points of other clusters.
Clustering has got immense applications in pattern
recognition, image analysis, bioinformatics and so on. In this
paper, the k-Means clustering algorithm has been applied in
customer segmentation. A MATLAB program (Appendix) of
the k-Means algorithm was developed, and the training was
realized using z-score normalized two-feature dataset of 100 Figure 4: Positions of the centroids and their cluster members
training patterns acquired from a retail business. After several after of k-Means algorithm
iterations, four stable clusters or customer segments were
identified. The two features considered in the clustering are The algorithm has a purity measure of 0.95 indicating 95%
the average amount of goods purchased by customer per accurate segmentation of the customers. Insight into the
month and the average number of customer visits per month. business’s customer segmentation will avail it with the
From the dataset, four customer clusters or segments were following advantages: the ability of the business to customize
identified and labeled thus: High-Buyers-Regular-Visitors market programs that will be suitable for each of its customer
(HBRV), High-Buyers-Irregular-Visitors (HBIV), Low segments; business decision support in terms of risky
Buyers-Regular-Visitors (LBRV) and Low-Buyers-Irregular situations such as credit relationship with its customers;
Visitors (LBIV). Furthermore, for any input pattern that was identification of products associated with each segments and
not in the training set, its cluster can be correctly extrapolated how to manage the forces of demand and supply; unraveling
by normalizing it and computing its similarities from the some latent dependencies and associations amongst
cluster centroids associated with each of the clusters. It will customers, amongst products, or between customers and
hence be assigned to any of clusters with which it has the products which the business may not be aware of; ability to
closest similarity. predict customer defection and which customers are most
likely to defect; and raising further market research questions
as well as providing directions to finding the solutions.

5. k-mean Algorithm in Pattern

Recognition[10]
Extracting the information fromimages and understanding
them such that the extractedinformation can be used for
several tasks is an importantcharacteristic of Machine
learning.Imagesegmentation is one of the initial steps in
direction ofunderstanding images and then finds the different
objects inthem.Image segmentation entails the separation or
division of theimage into areas of similar attributes. It is one
of the most crucial components of image analysis andpattern
recognition and still is considered as most challengingtasks
for the image processing and image analysis. It hasapplication
in several areas like Analysis of Remotely SensedImage,
Figure 3: The initialization stage Medical Science, Traffic System Monitoring, andFingerprint
Recognition and so on.The segmentation
methodincorporating clustering approaches encounters
greatdifficulties when computing the number of clusters that
arepresent in the feature space or extracting the
appropriatefeature. This type of image segmentation is widely
used due tothe simplicity of understanding and more accurate
result.There are different methods and one of the most
popular methods is k-means clustering algorithm. K-means
clustering algorithm is an unsupervised algorithm and it is
used to segment the interest area from the background.here
some examples for using this algorithm in pattern recognition.

Volume 6 Issue 8, August 2017

www.ijsr.net
Licensed Under Creative Commons Attribution CC BY
Paper ID: ART20176024 DOI: 10.21275/ART20176024 1581
 International Journal of Science and Research (IJSR)
ISSN (Online): 2319-7064
Index Copernicus Value (2015): 78.96 | Impact Factor (2015): 6.391
5.1 using k-mean algorithm to detect an infected part of
fruit[11] The basic aim of the proposed approach is to segmentcolors
automatically using the K-means clustering techniqueand
Plenty fruits are imported from theother nations such as L*a*b* color space. The introduced framework of
oranges, apples etc. Manual identificationof defected fruit is defectsegmentation operates in six steps as follows:
very time consuming. anovel defect segmentation of fruits
based on color features withK-means clustering unsupervised Step 1. Read the input image of defected fruits.
algorithm is presented.Colorimages of fruits for defect Step 2. Transform Image from RGB to L*a*b* ColorSpace.
segmentation is used. Defect segmentation iscarried out into We have used L*a*b* color spacebecause it consists of a
two stages. At first, the pixels are clustered basedon their luminosity layer in 'L'*channel and two chromaticity layer in
color and spatial features, where the clustering process 'a*' and'b*’ channels. Using L*a*b* color space
isaccomplished. Then the clustered blocks are merged to a iscomputationally efficient because all of the
specificnumber of regions. Using this two step procedure, it is colorinformation is present in the 'a*' and 'b*' layersonly.
possible toincrease the computational efficiency avoiding Step 3. Classify Colors using K-Means Clustering in'a*b*'
feature extractionfor every pixel in the image of fruits. Space. To measure the difference betweentwo colors,
Although the color is notcommonly used for defect Euclidean distance metric is used.
segmentation, it produces a highdiscriminative power for Step 4. Label Each Pixel in the Image from the Resultsof K-
different regions of image. Thisapproach thus provides a Means. For every pixel in our input, K-meanscomputes an
feasible robust solution for defectsegmentation of fruits. taken index corresponding to acluster. Every pixel of the image will
apple is taken as a case study andevaluated the proposed be labeledwith its cluster index.
approach using defected apples. Theexperimental results Step 5. Generate Images that Segment the Input Imageby
clarify the effectiveness of proposedapproach to improve the Color. We have to separate the pixels inimage by color using
defect segmentation quality in aspects ofprecision and pixel labels, which willresult different images based on the
computational time. The simulation results revealthat the number ofclusters.Programmatically determine the index of
proposed approach is promising.Image segmentation using k- each clustercontaining the defected part of the fruit because
means algorithm is quite usefulfor the image analysis. An K-means doesnot return the same cluster index value every
important goal of imagesegmentation is to separate the object time. But we cando this using thecenter value of clusters,
and background clearregardless the image has blur boundary. which contains themean value of 'a*' and 'b*' for each cluster.
Defect segmentationof fruits can be seen as an instance of
image segmentation inwhich number of segmentation is not
clearly known.

Figure 5: K-Means clustering for an apple fruit that is infected with apple scab disease with four clusters (a) The infected fruit
image, (b) first cluster, (c) second cluster, (d) third cluster, and (e) fourth cluster, respectively, (f) single gray-scale image
colored based on their cluster index.

Volume 6 Issue 8, August 2017

www.ijsr.net
Licensed Under Creative Commons Attribution CC BY
Paper ID: ART20176024 DOI: 10.21275/ART20176024 1582
 International Journal of Science and Research (IJSR)
ISSN (Online): 2319-7064
Index Copernicus Value (2015): 78.96 | Impact Factor (2015): 6.391

Figure 6: Defect segmentation results of apples (a) Images before segmentation, (b) Images after segmentation.

Experimental results suggest that the proposed approach

isable toaccurately segment the defected area of fruits
presentin the image. K-means based defect segmentation
approach isalso segment defected area with the stem and
calyx of thefruits.

5.2 using K-Means Clustering for Leukemia Image

Segmentation[12]

During the unfolding measures that are taken for the purpose
of leukemia detection, segmentation of blood cells is a vital
step. There are several method can used for solving this stat
.k-mean algorithm method is used and the results show that
the segmentation based on K-means clustering gives good
results.

The results show that thesegmentation using k-means

clustering was implemented successfully as shown in fig.(7).
The significant changes between segmented WBC (White
Blood Cell) in and background in leukemia images can
easily be seen. The fully segmented WBC (White Blood
Cell) is achieved by application of the algorithm. And better
results were obtained when K-Means Clustering was used
compering with another algorithm result.

Figure 7: (a) Original image and resultant image after

applying : (b) Linear Contrast, (c) HSI color Model, (d) K-
Means Clustering, (e) Median Filter (f) Binary image and (g)
Final segmented image.

6. Advantage and disadvantage of k-mean

algorithm
6.1 Advantages

This clustering methodology which we described earlier, has

some benefits comparing to others. The most important ones
are:

Volume 6 Issue 8, August 2017

www.ijsr.net
Licensed Under Creative Commons Attribution CC BY
Paper ID: ART20176024 DOI: 10.21275/ART20176024 1583
 International Journal of Science and Research (IJSR)
ISSN (Online): 2319-7064
Index Copernicus Value (2015): 78.96 | Impact Factor (2015): 6.391
 Lots of Applications – It has several live world [6] A. Mahendiran, N. Saravanan, N. Venkata Subramanian
implementations on many different subjects. We talk And N. Sairamm, Implementation Of K-Means
about the more relevant later in this article. Clustering In Cloud Computing Environment, Research
 Fast – Achieves the final result of its iterations in a fast Journal Of Applied Sciences, Engineering And
way due to the simplicity of the algorithm. Technology 4(10): 1391-1394, 2012.
 Simple and reliable – The process is fairly simple and [7] Zhexue Huang ,A Fast Clustering Algorithm To Cluster
always terminates, solving the problem with a solution set Very Large Categorical Data Sets InData Mining,
even for large data sets of information. Cooperative Research Centre For Advanced
 Efficient – This method presents a good solution with Computational Systems Csiro Mathematical And
relative low computational complexity for the clustering Information Sciences, Established Under The Australian
problem. Government’s Cooperative Research Centers Program.
 Good Solutions – Provides the best result set specifically [8] RuiXu,Survey Of Clustering Algorithms, IEEE
when data points are fairly separated. Transactions On Neural Networks, Vol. 16, No. 3,
May 2005 .
6.2 Disadvantages [9] Chinedu Pascal Ezenkwu, Simeon Ozuomba,
Constance Kalu, Application Of K-Means Algorithm
However this implementation has some problems which For Efficient Customer Segmentation: A Strategy For
need to be addressed. We provide you a list of the major Targeted Customer Services, (Ijarai) International
ones: Journal Of Advanced Research In Artificial
 No Categorical Data – One of the bigger problems of k- Intelligence, Vol. 4, No.10, 2015.
means clustering is that it can’t be used on data entries [10] B.S.Vamsi Krishna, P.Satheesh, Suneel Kumar R.,
that can’t simulate a mean function. “Comparative Study Of K-Means And Bisecting K-
Means Techniques In Wordnet Based Document
 Set Number of Clusters – In this algorithm the number of
Clustering”, International Journal Of Engineering And
partitions must be pre-defined. If this number is badly set,
Advanced Technology, Volume-1, Issue-6, August
the implementation and results will suffer a lot. Therefore
you should use techniques to estimate the number of 2012
[11] Shiv Ram Dubey1, Pushkar Dixit2, Nishant Singh3, Jay
clusters.
Prakash Gupta4,Infected Fruit Part Detection Using K-
 Result Set – As we explained before, the result set of this
Means Clustering Segmentation
process might not be optimal.
Technique,International Journal Of Artificial
 Initialization Method – Depending on the chosen
Intelligence And Interactive Multimedia, Vol. 2, Nº 2.,
initialization process, the results will differ.
[12] MashiatFatma, Jaya Sharma,Leukemia Image
Segmentation Using K-Means Clustering And Hsi
7. Conclusion Color Image Segmentation,International Journal Of
Computer Applications (0975 – 8887) Volume 94 – No
By studying several application of k-means algorithm in 12, May 2014.
both side data mining and pattern recognition where its use
as clustering method with data mining giving a promised
results and using as segmentation result when it used in
pattern recognition and also give a very good result ,so that's
mean it's an efficient algorithm in both state .

References
[1] John A. Hartigan , Clustering Algorithms, John Wiley &
Sons New York , London , Sydney , Toronto,1975 .
[2] Fayyad, U.M., G. Piatetsky Shapiro, P. Smyth And R.
Uthurusamy, Advances In Knowledge Discovery And
Data Mining,Aaai Press/The Mit Press, Pp: 573-592,
1996.
[3] M.Jianliang, S.Haikun And B.Ling, “The Application
On Intrusion Detection Based On K-Means Cluster
Algorithm”, IEEE International Conference On
Information Technology And Applications, 2009.
[4] Lei Li, De-Zhang Yang, Fang-Cheng Shen,A Novel
Rule-Based Intrusion Detection System Using Data
Mining,978-1-4244-5539, IEEE,2010.
[5] Alan Jose, S. Ravi And M. Sambath, Brain Tumor
Segmentation Using K-Means Clustering And Fuzzy C-
Means Algorithm And Its Area Calculation. In
International Journal Of Innovative Research In
Computer And Communication Engineering, Vol. 2,
Issue 2, March ,2014.

Volume 6 Issue 8, August 2017

www.ijsr.net
Licensed Under Creative Commons Attribution CC BY
Paper ID: ART20176024 DOI: 10.21275/ART20176024 1584