International Journal of Computer Applications (0975 – 8887)

Volume 68– No.10, April 2013

Clustering Algorithm for Spatial Data Mining: An

Overview
A.Padmapriya, N.Subitha
M.C.A.,M.Phil.,Ph.D Research scholar
Department of Computer Science and Engineering Department of Computer Science and Engineering
Alagappa University Alagappa University, Karaikudi
Karaikudi

ABSTRACT  forecasting-discovering patterns from which one

Spatial data mining practice for the extraction of useful can make reasonable predictions regarding future
information and knowledge from massive and complex spatial activities, such as the prediction that people who
database. Most research in this area has focused on efficient join an athletic club may take exercise classes
clustering algorithm for spatial database to analyze the
complexity. This paper introduces an active spatial data Data mining has been popularly treated as a synonym of
mining approach that extends the current spatial data mining knowledge discovery in database although some researchers
algorithms to efficiently support user-defined triggers on view data mining as an essential step of knowledge discovery.
dynamically evolving spatial data. It shows that spatial data
In general, a knowledge discovery process consists of an
mining is a promising field, with fruitful research results and
iterative sequence of the following step
many challenging issues.
1. Data cleaning, which handles noisy, erroneous,
Keywords missing, or irrelevant data.
Spatial data mining, Spatial database, K-mean, Spatial
relationship, Datamining. 2. Data integration, where multiple, heterogeneous data
source may be integrated into one.
1. INTRODUCTION
Data mining is the process of discovering interesting, 3. Data selection, where data relevant to the analysis
knowledge such as patterns, associations, changes, anomalies task are retrieved from the database.
and significant structures, from large amount of data stored in
database, data warehouses or other information repositories 4. Data transformation, where data are transferred or
[6]. Due to the wide availability of huge amounts of data in consolidated into forms appropriate for mining by performing
electronic forms, the imminent need for turning such data into summary or aggregation operations.
useful information and knowledge for broad application
including market analysis, business management, and decision 5. Data mining, which is an essential process where
support, data mining has attracted a great deal of attention in intelligent methods are applied in order to extract data
information industry in recent years[7]. Data mining can be patterns.
performed on data represented in quantitative, textual, or
multimedia forms. Data mining applications can use a variety 6. Pattern evaluation which is to identify the truly
of parameters to examine the data. They include interesting patterns representing knowledge based on some
interestingness measures.
 association-patterns where one event is connected to
another event, such as purchasing a pen and 7. Knowledge presentation, where visualization and
purchasing paper, knowledge representation technique are used to present the
mined knowledge to the user.
 sequence or path analysis -patterns where one event
leads to another event, such as the birth of a child With the widely available relational database system
and purchasing diapers, and data warehouses, the four processes: data cleaning, data
integration, data selection, and data transformation, can be
 classification-identification of new patterns, such as performed by constructing data warehouses and performing
coincidences between duct tape purchases and some OLAP operations on the constructed data warehouses.
plastic sheeting purchases, The data mining, pattern evaluation and knowledge
presentation processes are sometimes integrated into one
 clustering-finding and visually documenting groups (possibly iterative) process, referred as data mining [5].
of previously unknown facts, such as geographic
location and brand preferences,

28
 International Journal of Computer Applications (0975 – 8887)
Volume 68– No.10, April 2013

Interpretation

Knowledge

Data mining

Patterns
Transformation

Preprocessing
Transformed Data

Preprocessed Data
Selection

Target Data

Data

Figure l: An overview of the steps comprising the KDD process

2. BACKGROUND STUDY A data mining system may accomplish one or more of the
In general, data mining tasks can be classified into two following data mining tasks [1, 4].
categories: descriptive data mining and predictive data
mining. The former describes the data set in a concise and 1. Class description. Class description provides a concise
summary manner and presents interesting general properties and succinct summarization of a data and distinguishes
of the data whereas the latter construct one or a set of models, it from others .The summarization of a collection of data
performs inference on the available set if data, and attempts to is called class characterization; whereas the comparison
predict the behavior of new data sets. between two or more collections of data is called class
comparison or discrimination. Class description should
cover not only its summary properties, such as count,

29
 International Journal of Computer Applications (0975 – 8887)
Volume 68– No.10, April 2013

sum, and average, but also its properties on data ensure that the inter-cluster similarity is low and the intra-
dispersion, such as variance, quartile, etc. cluster similarity is high [10]. For example, one may cluster
the houses in an area according to their house category, floor
For example, class description can be used to compare area, and geographical locations.
European versus Asian sales of a company, identify the
important factors which discriminate the two classes, and Data mining research has been focused on high quality and
present a summarized overview. scalable clustering methods for large databases and
multidimensional data warehouses.
2. Association. Association is the discovery of association
relationships or correlations among a set of items. They 6. Time-series analysis. Time-series analysis is to analyze
are often expressed in the rule form showing attribute- large set of time-series data to find certain regularities and
value conditions that occur frequently together in a interesting characteristics, including search for similar
given set of data. An association rule in the form of sequence or subsequence patterns, periodicities, trends and
X→Y is interpreted as “database tuples that satisfy X deviations. For example, one may predict the trend of the
are likely to satisfy Y”. stock values for a company based on its stock history,
Association analysis is widely used in transaction data business situation, competitor’s performance, and current
analysis for directed marketing, catalog design, and market.
other business decision making process.
There are also other data mining tasks, such as outlier
Substantial research has been performed recently on analysis, etc. Identification of new data mining tasks to make
association analysis with efficient algorithms proposed, better use of the collected data itself is an interesting research
including the level-wise Apriori search, mining multiple-level, topic.
multi-dimensional associations, mining associations for
numerical, categorical, and interval data, meta-pattern directed Applications
or constraint-based mining, and mining correlations.
Data mining is a young discipline with wide and diverse
3. Classification. Classification analyzes a set of training data applications, there is still a nontrivial gab between general
(i.e., a set of objects whose class label is known) and principles of data mining tools for particular applications.
constructs a model for each class based on the features in the
data. A decision tree rules is generated by such a classification 1. Biomedical and DNA Data Analysis.
process, which can be used for better understanding of each
2. Financial Data Analysis.
class in the database and for classification of future data [1].
For example, one may classify diseases and help predict the 3. Retail Industry.
kind of diseases based on the symptoms of patients.
4. Telecommunication Industry.
There have been many classification methods developed in
the fields of machine learning, statistics, database, neural
3. SPATIAL DATA MINING
network, rough sets, and others. Classification has been used
in customer segmentation, business modeling, and credit Spatial data are the data that have spatial or location
analysis. component, and they the information, which is more complex
than classical data. A spatial database stores spatial data
4. Prediction. This mining function predicts the possible represents by spatial data types and spatial relationship and
values of some missing data or the value distribution of among data [6, 8].
certain attributes in a set of objects. It involves the finding of Spatial data is a highly demanding field because huge
the set of attributes relevant to the attribute of interest (e.g., by amounts of spatial data have been collected in various
applications, ranging from remote sensing, to geographical
some statistical analysis) and predicting the value distribution
information systems (GIS), computer cartography,
based on the set of data o f data similar to the selected objects. environmental assessment and planning[8] etc.
For example, an employee’s potential salary can be predicted
based on the salary distribution of similar employees in the Data Attributes
company. Usually, regression analysis, generalized linear
model, correlation analysis and decision trees are useful tools DATA = the (WHAT) dimension determines an attribute of
in quality prediction. Genetic algorithms and neural network an object.
models are also popularly used in prediction. SPATIAL DATA = (WHERE) & (WHAT) denotes attribute
data referenced to a specific location.
The Attributes of spatial objects are highly dependent on
5. Clustering. Clustering analysis is to identify clusters
location and often influenced by neighboring objects.
embedded in the data, where a cluster is a collection of data
objects that are “similar” to one another. Similarity can be
expressed by distance functions, specified by users or experts.
A good clustering method produces high quality clusters to

30
 International Journal of Computer Applications (0975 – 8887)
Volume 68– No.10, April 2013

Spatial Database (C) Applications

A spatial database stores a large amount of space-related Some of the applications of spatial data mining are listed
data, such as maps, preprocessed remote sensing or medical below,
imaging data, and VLSI chip layout data. Spatial database
carry topological and or distance information, usually  Geographic information systems,
organized by sophisticated, multidimensional spatial indexing
structures that are accessed by spatial data access methods and
 Geo marketing
often require spatial reasoning, geometric computation, and  remote sensing
spatial knowledge representation techniques [12].  image database exploration
 medical imaging
Spatial Data mining  navigation
 traffic control
Spatial data mining is the process [18] of discovering  environmental studies
interesting and previously un-known, but potentially useful
patterns from large spatial datasets. Extracting interesting and
useful patterns from spatial datasets is more difficult than (D) Clustering Methods
extracting the corresponding patterns from traditional numeric
and categorical data due to the complexity of spatial data The collection of clusters is known as clustering.
types, spatial relationships, and spatial autocorrelation. Spatial Goal: like Generalization, to reveal relationships between
data mining , i.e., mining knowledge from large amounts of spatial and non-spatial attributes
spatial data, is a highly demanding field because huge
amounts of spatial data have been collected in various There are various types of clustering as follows
applications, ranging from remote sensing, to geographical
information system(GIS), computer cartography, 1. Hierarchical Methods
environmental assessment and planning[8] etc. The collected
data far exceeded human’s ability to analyze. Recent studies It can have two types of algorithms they are [9],
on data mining have extended the scope of data mining from  Agglomerative Algorithm
relational and transactional databases to spatial databases.
 Divisive Algorithm
(A) Spatial Data Mining Methods 2. Partitioning Methods

Spatial data mining has to perform various methods some of It can contain many types of algorithms they are [10],
them are mentioned below  Nearest Neighbor Algorithm
 Density Based Algorithm
1. Generalization Based Knowledge Discovery
2. Clustering Methods  K-Medoids Methods
3. Aggregate Proximity Measuring  K-Mean Methods
4. Spatial Association Rules
3. Grid Based Methods
Among the four methods the research is based on
clustering method. 4. Methods Based on Co-occurrence of Categorical Data.

Goals 5. Density Based methods.

There are different goals of spatial data mining are

ordered below,
4. ENHANCED K-MEANS ALGORITHM
ON SPATIAL DATASET
 Understanding spatial data
 Discovering spatial relationships and relationships K-Means algorithm introduced by J.B. Mac Queen in 1967,
between spatial and non-spatial data is one of the most common clustering algorithms and it is
 Constructing spatial knowledge bases considered as one of the simplest unsupervised learning
 Reorganizing spatial databases algorithms that partition feature vectors into k clusters so that
 Optimizing spatial queries the within group sum of squares is minimized.

There are several variants of the k-means clustering

(B)Challenges of Spatial Data Mining
algorithm, but most variants involve an iterative scheme that
operates over a fixed number of clusters, while attempting to
Spatial Data mining must efficiently overcome the following satisfy the following properties: Each class has a center which
challenges [11, 14]: is the mean position of all the samples in that class.
1. A crucial challenge to spatial data mining is the
exploration of efficient spatial data mining techniques.
2. Huge amount of spatial data.
3. Complexity of spatial data types and spatial access
methods.

31
 International Journal of Computer Applications (0975 – 8887)
Volume 68– No.10, April 2013

measure of distance or similarity like the Euclidean Distance

PROCEDURE OF K-MEAN ALGORITHM Measure or Manhattan/City-Block Distance Measure.

Step 1: Place randomly initial group centroids into the We have to re-assigns each record in the dataset to the most
similar cluster and re-calculate the arithmetic mean of all the
2d space.
clusters in the dataset. The arithmetic mean of a cluster is the
arithmetic mean of all the records in that cluster.
Step 2: Assign each object to the group that has the
closest centroid. For Example, if a cluster contains two records where the
record of the set of measurements for
Step 3: Recalculate the positions of the centroids.
John = {20, 170, 80} and
Step 4: If the positions of the centroids didn't change Henry = {30, 160, 120},

go to the next step, Then the arithmetic mean P mean is represented as

Else go to Step 2. P mean = {Age mean, Height mean, Weight mean).

Step 5: End Age mean = (20 + 30)/2,

Height mean = (170 + 160)/2 and
Weight mean = (80 + 120)/2.

The arithmetic mean of this cluster = {25, 165, 100}.

(A) Working
It accepts the number of clusters to group data into, and the
This new arithmetic mean becomes the center of this new
dataset to cluster as input values. It then creates the first K
cluster. Following the same procedure, new cluster centers are
initial clusters (K= number of clusters needed) from the
formed for all the existing clusters.
dataset by choosing K rows of data randomly from the dataset.
It K-Means re-assigns each record in the dataset to only one
For Example, if there are 10,000 rows of data in the dataset
of the new clusters formed. A record or data point is assigned
and 3 clusters need to be formed, then the first K=3 initial
to the nearest cluster (the cluster which it is most similar to)
clusters will be created by selecting 3 records randomly from
using a measure of distance or similarity like the Euclidean
the dataset as the initial clusters [14, 15]. Each of the 3 initial
Distance Measure or Manhattan/City-Block Distance
clusters formed will have just one row of data.
Measure. The preceding steps are repeated until stable clusters
are formed and the K-Means clustering procedure is
The K-Means algorithm calculates the Arithmetic Mean of
completed [17]. Stable clusters are formed when new
each cluster formed in the dataset. The Arithmetic Mean of
iterations or repetitions of the K-Means clustering algorithm
a cluster is the mean of all the individual records in the
does not create new clusters as the cluster center or Arithmetic
cluster. In each of the first K initial clusters, there is only one
Mean of each cluster formed is the same as the old cluster
record [16]. The Arithmetic Mean of a cluster with one record
center. There are different techniques for determining when a
is the set of values that make up that record.
stable cluster is formed or when the k-means clustering
algorithm procedure is completed.
For Example if the dataset we are discussing is a set of
Height, Weight and Age measurements for students in a
(A) Computational complexity
University, where a record P in the dataset S is represented by
NP-hard in general Euclidean space d even for 2 clusters.
a Height, Weight and Age measurement, then
NP-hard for a general number of clusters k even in the plane.
If k and d are fixed, the problem can be exactly solved in
P = {Age, Height, Weight).
time O (n dk+1 log n), where n is the number of entities to
be clustered.
Then a record containing the measurements of a student John,
would be represented as
It has some of the advantages are relatively efficient: O (tkn),
where n is the number of instances, c is the number of
John = {20, 170, 80}
clusters, and t is the number of iterations. Normally, k, t << n.
Often terminates at a local optimum. The global optimum may
Where
be found using techniques such as: simulated annealing or
John's Age = 20 years,
genetic algorithms
Height = 1.70 meters and
Weight = 80 Pounds.
Also has some disadvantages it’s applicable only when mean
is defined.
Since there is only one record in each initial cluster then the
 Need to specify c, the number of clusters, in
Arithmetic Mean of a cluster with only the record for John as
advance.
a member = {20, 170, 80}.
 Unable to handle noisy data and outliers.
 Not suitable to discover clusters with non-convex
It Next, K-Means assigns each record in the dataset to only
shapes.
one of the initial clusters. Each record is assigned to the
nearest cluster (the cluster which it is most similar to) using a

32
 International Journal of Computer Applications (0975 – 8887)
Volume 68– No.10, April 2013

5. CONCLUSION [9] G. Karypis, E.-H. Han, and V. Kumar, “CHAMELEON:

Data mining/ Knowledge Discovery of spatial Data is a large, A Hierarchical Clustering Algorithm Using Dynamic
active field of research with wide application in GIS, remote Modeling,” Computer, vol. 32, no. 8, pp 68–75, Aug.
sensing, medical imaging, traffic control, environmental 1999
studies etc. Although, the field is quite young, a number of [10] L. Kaufman and P.J. Rousseeuw, Finding Groups in
algorithms and techniques have been proposed to discover Data: an Introduction to Cluster Analysis.John Wiley &
various kinds of knowledge from spatial data with the help of Sons, 1990.
K-means clustering algorithm. This work motivated us and
gives future direction towards designing an efficient [11] Koperski K. Adhikary J., Han J. 1996 “Knowledge
clustering algorithm for spatial database with reduced Discovery in Spatial Databases: Progress and
complexity. The variety of yet unexplored topics and Challenges”, Proc. SIGMOD Workshop on Research
problems makes knowledge discovery in spatial database an Issues in Data Mining and Knowledge Discovery,
attractive and challenging research field. Technical Report 96-08, University of British Columbia,
Vancouver,Canada.
[12] K. Koperski and J. Han. Discovery of Spatial Association
6. REFERENCES Rules in Geographic Information Databases. In Proc. th
[1] R. Agrawal, M. Mehta, J. Shafer, R. Srikant, A. Arning,
Int’l Symp. On Large Spatial Databases (SSD ‘95), pp.
T. Bollinger. The Quest Data Mining System.
47 66, Portland, Maine, August 1995
Proceedings of 1996 International Conference on Data
Mining and Knowledge Discovery(KDD’96), Portland, [13] Krzysztof Koperski, Junas Adhikary, JiaweiHan. Spatial
Oregon, pp. 244-249, August 1996. Data Mining: Progress and Challenges Survey Paper.
Workshop on Research Issues on Data Mining and
[2] K. Alsabti, S. Ranka, and V. Singh, ªAn Efficient k-
Knowledge Discovery, 1996
means Clustering Algorithm,º Proc. First Workshop High
Performance Data Mining, Mar. 1998 [14] G. Milligan and M. Cooper, “An Examination of
Procedures for Determining the Number of Clusters in a
[3] P. S. Bradley, U. Fayyad, and C. Reina, "Scaling
Data Set,”Psychometrika, vol. 50, pp. 159–179, 1985
Clustering Algorithms to Large Databases", Proc. 4 th
International Conf. on Knowledge Discovery and Data [15] Paul S. Bradley and Usama M. Fayyad. Refining initial
Mining (KDD-98). AAAI Press, Aug. 1998 points for k-means clustering. In Jude W. Shavlik,
editor, ICML, pages 91–99. Morgan Kaufmann, 1998.
[4] M. S. Chen, J. Han, and P.S.Yu. Data Mining: An
Overwiew from a Database Perspective. IEEE [16] Raymond T. Ng and Jiawei Han, CLARANS: A Method
Transcations on Knowledge and Data Engineering, for Clustering Objects for Spatial Data Mining, IEEE
8(6):883, 1996. TRANSACTIONS ON KNOWLEDGE and DATA
ENGINEERING, Vol. 14, No. 5,
[5] Dan Pelleg and Andrew W. Moore. Accelerating exact
k-means algorithms with geometric reasoning. In KDD, [17] Shai Ben-David, D´avid P´al, and Hans Ulrich Simon.
pages 277–281, 1999. Stability of k-means clustering. Lecture Notes in
Computer Science, 4539:20–34, 2007
[6] Ester M., Kriegel H.-P., and Sander J. 1997 “Spatial Data
Mining: A Database Approach”, Proc. 5th Int. Symp. on [18] Shekhar, S., and Chawla, S. 2003. Spatial Databases A
Large Spatial Databases, Berlin, Germany, pp. 47-66. Tour. Prentic e Hall (ISBN 0-7484-0064-6).
[7] U. M. Fayyades, G. Piatetsky-Shapiro, P. Smyth, and R. [19] Tapas Kanungo, David M. Mount, Nathan S. Ne-
Uthurusamy (Eds). Advances in Knowledge Discovery tanyahu, Christine D. Piatko, Ruth Silverman, and An-
and Data Mining. AAAI/MIT Press, 1996. gela Y. Wu. An efficient k-means clustering algorithm
Analysis and implementation. IEEE Trans. Pattern Anal.
[8] W. Lu, J. Han, and B. C. Obi. Discovery of General
Mach. Intell., 24(7):881–892, 2002.
Knowledge in Large Spatial Databases. In Proc. Far East
Workshop on Geographic Information Systems pp. 275-
289, Singapore, June 1993

33