Introduction to Data Mining Techniques

Dr. Rajni Jain

1 Introduction
The last decade has experienced a revolution in information availability and
exchange via the internet. In the same spirit, more and more businesses and
organizations began to collect data related to their own operations. While the
database technologists have been seeking efficient means of storing, retrieving
and manipulating data, the machine learning community has focussed on
developing techniques for learning and acquiring knowledge from the data. At
times the data can be considered to be a gold mine for strategic planning for
research and development in this area which is often referred to as Data Mining
(DM) and Knowledge Discovery in Databases (KDD).
We are overwhelmed with data. People have been seeking patterns in data
since human life began. Hunters seek patterns in animal migration behaviour,
farmers seek patterns in crop growth, and politicians seek patterns in voter
opinion. A scientists job is to make sense out of data, to discover the underlying
model that governs the functioning of the physical world and encapsulate the same
in theories that can be used for predicting the future. As the background of all
scientific discoveries especially theories has been same, what is new about Data
Mining (DM)? The simple answer is that, in DM the volume of the stored data is
in the digital form and the search is automated or augmented by a computer. In
DM, it is important to understand the difference between a model and a pattern.
Model is a global summary of the dataset and makes statements about any point in
the full measurement space while pattern describes a structure, relationship to a
relatively small part of the data or the space in which the data would occur
[HMS01]. In 1960s, computers were increasingly applied to data analysis
problems and it was noted that if one searches long enough, one can always find
some model to fit in the dataset but complexity and size of the model were
important considerations. Also the aim is to generalize beyond the available data.
Figure 1 shows the history of databases systems and DM [HK01]. And Figure 2
presents the scope of data mining in KDD. Fayyad [Fay96] defined DM as a
process of finding models, interesting trends or patterns in large datasets in order
to guide decisions about future activities. It requires tools that can help in

explaining the data and which are also capable to make predictions out of that.
The data takes the form of a set of examples and the output takes the form of
predictions on the new examples.
2 Issues in Data mining
Data mining has evolved into an important and active area of research
because of the theoretical challenges and practical applications associated with the
problem of discovering interesting and previously unknown knowledge from realworld databases. The main challenges to the data mining and the corresponding
considerations in designing the algorithms are as follows:
1. Massive datasets and high dimensionality.
2. Overfitting and assessing the statistical significance.
3. Understandability of patterns.
4. Non-standard incomplete data and data integration.
5. Mixed changing and redundant data.
3 Tasks of Data Mining
Data mining as a term used for the specific set of six activities or tasks as follows:
1. Classification
2. Estimation
3. Prediction
4. Affinity grouping or association rules
5. Clustering
6. Description and visualization
The first three tasks - classification, estimation and prediction are all
examples of directed data mining or supervised learning. In directed data mining,
the goal is to use the available data to build a model that describes one or more
particular attribute(s) of interest (target attributes or class attributes) in terms of
the rest of the available attributes. The next three tasks association rules,
clustering and description are examples of undirected data mining i.e. no attribute
is singled out as the target; the goal is to establish some relationship among all the
attributes.

Data Collection and Database

Creation
(1960s and Earlier)
-Primitive File Processing

Database Management Systems

(1970s-early 1980s)
Hierarchical and network database
systems
Relational database systems
Data modelling tools
Query Languages:/SQL etc.
Query Processing
Online Transaction Processing (OLTP)

Advanced Databases Systems

(Mid-1980s-present)
-Advances data models:
extended relational, object
oriented, object relational,
deductive
-Application oriented:
Spatial, temporal, multimedia,
active, scientific, knowledge

Web-based Databases
Systems
(1990-present)
-XML-based database
systems
-Web Mining

Data Warehousing and Data Mining

(Late 1980s-present)
-Data warehouse and OLAP technology
-Data mining and knowledge discovery

New Generation of Integrated Information Systems

(2000-)

Figure 1: The history of databases systems and data mining

Cleaning and
Integration

Databases

Data
Warehouse

Selection
and
Transformation

Flat Files

Data
Mining

Knowledge

Evaluation
and
Presentation

Figure 2: Data mining as a step in the process of knowledge

discovery

3.1 Classification
Classification consists of examining the features of a newly presented object and
assigning to it a predefined class. The classification task is characterized by the
well-defined classes, and a training set consisting of preclassified examples. The
task is to build a model that can be applied to unclassified data in order to classify
it. Examples of classification tasks include:

Classification of credit applicants as low, medium or high risk

Classification of mushrooms as edible or poisonous

Determination of which home telephone lines are used for internet access

3.2 Estimation
Estimation deals with continuously valued outcomes. Given some input data, we
use estimation to come up with a value for some unknown continuous variables
such as income, height or credit card balance. Some examples of estimation tasks
include:

Estimating the number of children in a family from the input data of mothers
education

Estimating total household income of a family from the data of vehicles in the
family

Estimating the value of a piece of a real estate from the data on proximity of
that land from a major business centre of the city.

3.3 Prediction
Any prediction can be thought of as classification or estimation. The difference is
one of emphasis. When data mining is used to classify a phone line as primarily
used for internet access or a credit card transaction as fraudulent, we do not expect
to be able to go back later to see if the classification was correct. Our
classification may be correct or incorrect, but the uncertainty is due to incomplete
knowledge only: out in the real world, the relevant actions have already taken
place. The phone is or is not used primarily to dial the local ISP. The credit card
transaction is or is not fraudulent. With enough efforts, it is possible to check.
Predictive tasks feel different because the records are classified according to some
predicted future behaviour or estimated future value. With prediction, the only

way to check the accuracy of the classification is to wait and see. Examples of
prediction tasks include:

Predicting the size of the balance that will be transferred if a credit card
prospect accepts a balance transfer offer

Predicting which customers will leave within next six months

Predicting which telephone subscribers will order a valueadded service such

as three-way calling or voice mail.
Any of the techniques used for classification and estimation can be

adopted for use in prediction by using training examples where the value of the
variable to be predicted is already known, along with historical data for those
examples. The historical data is used to build a model that explains the current
observed behaviour. When this model is applied to current inputs, the result is a
prediction of future behaviour.
3.4 Association Rules
An association rule is a rule which implies certain association relationships among
a set of objects (such as occur together or one implies the other) in a database.
Given a set of transactions, where each transaction is a set of literals (called
items), an association rule is an expression of the form X Y , where X and Y are
sets of items. The intuitive meaning of such a rule is that transactions of the
database which contain X tend to contain Y. An example of an association rule is:
30% of farmers that grow wheat also grow pulses; 2% of all farmers grow both
of these items. Here 30% is called the confidence of the rule, and 2% the support
of the rule. The problem is to find all association rules that satisfy user-specified
minimum support and minimum confidence constraints.
3.5 Clustering
Clustering is the task of segmenting a diverse group into a number of similar
subgroups or clusters. What distinguishes clustering from classification is that
clustering does not rely on predefined classes. In clustering, there are no
predefined classes. The records are grouped together on the basis of self
similarity. Clustering is often done as a prelude to some other form of data mining
or modelling. For example, clustering might be the first step in a market

segmentation effort, instead of trying to come up with a one-size-fits-all rule for

determining what kind of promotion works best for each cluster.
3.6 Description and Visualization
Data visualization is a powerful form of descriptive data mining. It is not always
easy to come up with meaningful visualizations, but the right picture really can be
worth a thousands association rules since the human beings are extremely
practiced at extracting meaning from visual scenes.
Knowledge discovery goals are defined by the intended use of the system.
There are two types of goals: (1) verification and (2) discovery. With verification,
the system is limited to verifying the users hypothesis. With discovery, the
system autonomously finds new patterns. The discovery goal is further divided
into prediction, where the system finds patterns for predicting the future behaviour
of some entities and description, where the system finds patterns for presentation
to a user in human understandable form.
4 DM Techniques
Each of the above problems is relevant to the derivation of useful knowledge from
collection of data. Therefore methods for solving these problems, developed in the
disciplines of statistics, machine learning, fuzzy sets, rough sets or hybrids are
also directly relevant to DM, in particular, to conceptual data exploration. Various
popular techniques of classification which have been in use pertaining to each
disciplines are introduced in this section.
4.1 Statistics
The problem of abstracting knowledge from data has been tackled by statisticians,
long before the first artificial intelligence papers were published. For example,
correlation analysis applies statistical tools for analyzing the correlation between
two or more variables. Cluster analysis offers methods for discovering clusters in
large set of objects described by vector of values. Factor analysis tries to point the
most important variables describing clusters. Some of the popular techniques that
are used for supervised classification tasks are Linear Discriminants, Quadratic
Discriminants, K-nearest Neighbour, Nave Bayes, Logistic Regression and
CART.

4.2 Machine Learning

Statistical methods have difficulty incorporating subjective, non quantifiable
information in their models. They also have to assume various distributions of
parameters and independence of attributes. Various studies have concluded that
machine learning produces comparable (and often better) predictive accuracy. Its
good performance as compared to statistical methods can be attributed to the fact
that it is free from parametric and structural assumptions that underlie statistical
methods. Another weakness of statistical approaches to data analysis is the
problem of interpreting the results. Some of the machine learning techniques are
mentioned below.
4.2.1 Neural Networks
Artificial neural networks are computational models composed of many non linear
processing elements arranged in a pattern similar to biological neuron networks. A
typical neural network has an activation value associated with each node and a
weight value associated with each connection. An activation function governs the
firing of nodes and the propagation of data through network connections in
massive parallelism. The network can also be trained with examples through
connection weight adjustments [TQT96].
4.2.2 Genetic Algorithms
Genetic algorithms are search algorithms based on mechanics of natural selection
and natural genetics [Gol89]. They combine survival of the fittest among string
structures with a structured yet randomized information exchange to form a search
algorithm with some of the innovative flair of human search. In every generation,
a new set of strings is created using bits and pieces of the fittest of the old; an
occasional new part is tried for good measure. While randomized, genetic
algorithms are no simple random walk. They efficiently exploit historical
information to speculate on new search points with expected improved
performance. A simple GA that yields good result, is composed of three operators
namely reproduction, crossover and mutation. GAs differ from more normal
optimization and search procedures in four ways:
1. GAs work with coding of parameter set, not the parameter themselves.
2. GAs search from a population of points, not a single point.

3. GAs use objective function information, not derivatives or other auxiliary

knowledge.
4. GAs use probabilistic transitional rules, not deterministic rules.
4.2.3 Support Vector Machines
SVMs are the learning machines that can perform binary classification and
regression estimation tasks. They are becoming increasingly popular as a new
paradigm of classification and learning because of two important factors. First,
unlike the other classification techniques, SVMs minimize the expected error
rather than minimizing the classification error. Second, SVMs employ the duality
theory of mathematical programming to get a dual problem that admits efficient
computational methods.
4.2.4 Decision Tree Induction
A DT is a classification scheme which generates a tree and a set of rules,
representing the model of different classes, from a given dataset. As per Hans and
Kamber [HK01], DT is a flow chart like tree structure, where each internal node
denotes a test on an attribute, each branch represents an outcome of the test and
leaf nodes represent the classes or class distributions. The top most node in a tree
is the root node. Figure 3 refers to DT induced for dataset in Table 1. We have the
following rules corresponding to the tree.
1. If Hair=blonde and Lotion=no then Sunburn=yes;
2. If Hair=blonde and Lotion=yes then Sunburn=no;
3. If Hair=red then Sunburn=yes;
4. If Hair=brown then Sunburn=no;
Table 1: Sunburn Dataset
ID

Hair

Height

Weight

Lotion

Sunburn

X1

blonde

average

light

no

yes

X2

blonde

tall

average

yes

no

X3

brown

short

average

yes

no

X4

blonde

short

average

no

yes

X5

red

average

heavy

no

yes

X6

brown

tall

heavy

no

no

X7

brown

average

heavy

no

no

X8

blonde

short

light

yes

no

Hair
blonde
red

brown

Lotion
yes
yes

no

no

no

yes

Figure 3: DT obtained by using ID3 algorithm on Sunburn data

4.3 Fuzzy Logic

Fuzzy logic, which may be viewed as an extension of classical logical systems,
provides an effective conceptual framework for dealing with the problem of
knowledge representation in an environment of uncertainty and imprecision
[Zad89]. Some of the essential characteristics of fuzzy logic relate to the
following:
1. In fuzzy logic, exact reasoning is viewed as a limiting case of approximate
reasoning.
2. In fuzzy logic everything is a matter of degree.
3. Any logical system can be fuzzified.
4. In fuzzy logic, knowledge is interpreted as a collection of elastic or
equivalently, fuzzy constraint on a collection of variables.

Summary of basic concepts and techniques underlying the application of fuzzy

logic to knowledge representation and description of number of examples relating
to its use as a computational system is provided in [Zad89]. Fuzzy logic in its pure
form is not a technique for classification but it has been a very useful concept in
many hybrid techniques for classification.
4.4 Rough Sets Techniques
RS theory deals with approximation of sets or concepts by means of binary
relations constructed from empirical data based on the notion of indiscernibility
and the inability to distinguish between objects. Such approximations can be said
to form models of our target concepts, and hence in its typical use, falls in under
the bottom up approach to model construction. Rough set applications to data
mining generally proceed along the following directions:
1. Decision rule induction from attribute value table
2. Data filtration by template generation - This mainly involves extracting
elementary blocks from data based on equivalence relation. Genetic
algorithms are also sometimes used in this stage for searching.
5 Summary
Data mining involves extracting useful rules or interesting patterns from historical
data. There are many data mining tasks each of them further has many techniques.
No free lunch theorem exists i.e. a single technique is not suitable for all kinds of
data for all types of domains. Sometimes, hybrid techniques have been observed
to perform better as compared to the pure ones.
References
[HMS01]
[HK01]
[Fay96]
[TQT96]
[Gol89]
[Zad89]

Hand, D., Mannila, H., Smyth, P., Principles of Data Mining, Prentice Hall
of India, 2001
Han, J., Kamber, M. Data Mining Concepts and Techniques, Morgan
Kaufmann Publisher, 2001
Fayyad, U., Data Mining and Knowledge Discovery: Making Sense out of
Data, IEEE Expert, Oct. 20-25, 1996
Tan, C. L., Quah, T. S. and Teh, H. H., An Artificial Neural Network that
models Human Decision making, IEEE Computer, 64-70, 1996
Goldberg, D.E. Genetic Algorithms in Search Optimization and Machine
Learning, Addison -Wesley, 1989
Zadeh, L. A., Knowledge Representation in Fuzzy Logic, IEEE TKDE,
1(1):89-99, 1989