Vtu 7TH Sem Cse/ise Data Warehousing & Data Mining Notes 10cs755/10is74

DATA WAREHOUSING AND DATA MINING
Subject Code: 10IS74/10CS755

Hours/Week: 04
Total Hours: 52
I.A. Marks : 25
Exam Hours: 03
Exam Marks: 100
PART A
UNIT 1
6 Hours
Data Warehousing: Introduction, Operational Data Stores (ODS), Extraction Transformation Loading (ETL), Data Warehouses.
Design Issues, Guidelines for Data Warehouse Implementation, Data Warehouse Metadata
UNIT 2
6 Hours
Online Analytical Processing (OLAP): Introduction, Characteristics of OLAP systems, Multidimensional view and Data cube,
Data Cube Implementations, Data Cube operations, Implementation of OLAP and overview on OLAP Softwares.
UNIT 3
6 Hours
Data Mining: Introduction, Challenges, Data Mining Tasks, Types of Data, Data Preprocessing, Measures of Similarity and
Dissimilarity, Data Mining Applications
UNIT 4
8 Hours
Association Analysis: Basic Concepts and Algorithms: Frequent Itemset Generation, Rule Generation, Compact Representation
of Frequent Itemsets, Alternative methods for generating Frequent Itemsets, FP Growth Algorithm, Evaluation of Association
Patterns
PART - B
UNIT 5
6 Hours
Classification -1 : Basics, General approach to solve classification problem, Decision Trees, Rule Based Classifiers, Nearest
Neighbor Classifiers.
UNIT 6
6 Hours
Classification - 2 : Bayesian Classifiers, Estimating Predictive accuracy of classification methods, Improving accuracy of
clarification methods, Evaluation criteria for classification methods, Multiclass Problem.
UNIT 7
8 Hours
Clustering Techniques: Overview, Features of cluster analysis, Types of Data and Computing Distance, Types of Cluster
Analysis Methods, Partitional Methods, Hierarchical Methods, Density Based Methods, Quality and Validity of Cluster Analysis
UNIT 8
6 Hours
Web Mining: Introduction, Web content mining, Text Mining, Unstructured Text, Text clustering, Mining Spatial and Temporal
Databases.
Text Books:
1. Pang-Ning Tan, Michael Steinbach, Vipin Kumar: Introduction to Data Mining, Addison-Wesley, 2005.
2. G. K. Gupta: Introduction to Data Mining with Case Studies, 3 rd Edition, PHI, New Delhi, 2009.
TABLE OF CONTENTS
UNIT-1: DATA WAREHOUSING
1-10
UNIT 2: ONLINE ANALYTICAL PROCESSING (OLAP)
11-22
UNIT 3: DATA MINING
23-40
UNIT 7: CLUSTERING TECHNIQUES
41-58
UNIT 8: WEB MINING/TEMPORAL & SPATIAL DATA MINING
59-68
DATA WAREHOUSING & DATA MINING
UNIT 1: DATA WAREHOUSING

INTRODUCTION
A large company might have the following systems:
Human resources(HR)
Financials
Billing
Sales leads
Web sales
Customer support
Such systems are called OLTP systems (OnLine Transaction Processing).
The systems are mostly relational database systems designed for transaction processing.
The performance of OLTP systems is usually very important,
since such systems are used to support users(i.e. staff) who provide service to customers.
The systems must be able to deal with
insert & update operations and
quickly answering queries
The systems are not designed to handle management-queries efficiently (since management-queries
are often complex, requiring many joins & aggregations).
Focus of operational-managers is on improving business-processes across various business-units (for
example: customer support, inventory, and marketing). To achieve this, they require:
a single sign-on path to the enterprise-information
a single version of the enterprise-information
a high level of data accuracy
a user-friendly interface to the information
easy sharing of data across business-units of enterprise
Following are some solutions to meet the needs of management-staff in an enterprise:
1) Managers pose queries of interest to a mediator-system that
decomposes each query into appropriate subqueries for the systems
obtains results from those systems and
then combines & presents the result to the user
This is called lazy (or on-demand) query processing.
Advantage: The user is provided with up-to-date information.
Disadvantage: The management-queries may generate a heavy load on some OLTP
systems which may degrade the performance of the systems.
2) Collect the most common queries that the managers ask and
then run them regularly and
finally have the results available when a manager poses one of those queries.
This approach is called the eager query processing.
Advantage: Provides quick response.
Disadvantage: The information may not be up-to-date.
3) This approach
involves creation of a separate database that only stores information that is of
interest to the management-staff and
involves the following 2 steps:
i) The information needs of management-staff are analyzed and the systems that
store some or all of the information are identified.
ii) The new database is then used to answer management-queries and the
OLTP systems are not accessed for such queries.
It is literally true that you can succeed best and quickest by helping others to succeed.
1

ODS
ODS stands for Operational Data Store.
This is defined as a subject-oriented, integrated, volatile, current-valued data store, containing only
corporate-detailed data.
ODS is subject-oriented i.e. it is organized around the major data-subjects of an enterprise
ODS is integrated i.e. it is a collection of data from a variety of systems to provide an
enterprise-wide view of the data
ODS is volatile i.e. data in the ODS changes frequently as new information refreshes ODS
ODS is current-valued i.e. it is up-to-date and reflects the current status of the information
ODS is detailed i.e. it is detailed enough to serve needs of management-staff in enterprise
Benefits of ODS to an enterprise:
1) An ODS is the unified-operational view of enterprise. It provides operational-managers
improved access to important operational-data. This view may assist in better understanding of
business & customer.
2) The ODS can be much more effective in generating current reports without having to access
the OLTP.
3) The ODS can shorten the time required to implement & populate a data warehouse system.
An ODS may be of following types:
1) An ODS that is essentially a reporting-tool for administrative purposes may be fairly simple.
Such an ODS is usually updated daily. (It provides reports about business transactions for that
day, such as sales totals or orders filled).
2) An ODS may be designed to track more complex information such as product & location
codes. Such an ODS is usually updated hourly.
3) An ODS may be designed to support CRM (Customer Relationship Management).
ODS DESIGN & IMPLEMENTATION
The extraction of information from source databases needs to be efficient (Figure 7.1).
The quality of data needs to be maintained.
Suitable checks are required to ensure quality of data after each refresh.
An ODS would be required to satisfy normal integrity constraints, for example,
existential- & referential-integrity
appropriate action to deal with nulls
An ODS is a read only database. Users should not be allowed to update information in ODS.
Populating an ODS involves an acquisition process of extracting, transforming & loading data from
source systems. This process is called ETL(Extraction, Transformation and Loading).
Checking for anomalies & testing for performance are necessary before an ODS system can go
online.
Its human nature not to appreciate all we have until it's lost.
2

WHY A SEPARATE DATABASE?
Q: Why an ODS should be separate from the operational-databases?
Ans: Because from time to time, complex queries are likely to degrade performance of OLTP
systems. The OLTP systems have to provide a quick response to operational-users as businesses
cannot afford to have response time suffer when a manager is running a complex query.
ZLE
ZLE stands for Zero Latency Enterprise.
This is used for near real-time integration of operational-data so that there is no significant delay in
getting information from one system to another system in an enterprise.
The heart of a ZLE system is an ODS.
A ZLE data store(ZDS) is something like an ODS that is integrated & up-to-date.
Aim of a ZLE data store is to:
allow management a single view of enterprise-information by bringing together relevant data
in real-time &
provide management a "360-degree" view of the customer
Characteristics of ZLE:
It has a unified view of the enterprise operational-data.
It has a high level of availability and it involves online refreshing of information.
Since a ZLE needs to support a large number of concurrent users (for example call centre users), a
fast turnaround time for transactions and 24/7 availability is required.
Q: How does a ZLE data store fit into enterprise-information architecture?
Ans: Either a ZLE data store can be used almost like an ODS that is refreshed in real-time or
a ZLE can take over some of the roles of the data-warehouse.
ETL
The ETL process involves extracting, transforming and loading data from multiple source-systems.
The process is much more complex and tedious. The process may require significant resources to
implement.
Different data-sources tend to have
different conventions for coding information &
different standards for the quality of information
Building an ODS requires data filtering, data cleaning and integration.
Data-errors at least partly arise because of unmotivated data-entry staff.
Successful implementation of an ETL system involves resolving following issues:
1) What are the source systems? These systems may include relational database systems,
legacy systems.
2) To what extent are the source systems and the target system interoperable? The more
different the sources and target, the more complex the ETL process.
3) What ETL technology will be used?
4) How big is the ODS likely to be at the beginning and in the long term? Database systems
tend to grow with time. Consideration may have to be given to whether some of the data
from the ODS will be archived regularly as the data becomes old and
is no longer needed in the ODS
5) How frequently will the ODS be refreshed or reloaded?
6) How will the quality and integrity of the data be monitored? Data cleaning will often
required to deal with issues like missing values, data formats, code values, primary keys
and referential integrity.
7) How will a log be maintained? A dispute may arise about the origin of some data. It is
therefore necessary to be able to not only log which information came from where but also
when the information was last updated.
8) How will recovery take place?
9) Would the extraction process only copy data from the source systems and not delete the
original data?
10) How will the transformation be carried out? Transformation could be done in either
source OLTP system, ODS or staging area.
11) How will data be copied from non-relational legacy systems that are still operational?
Selfless service is the rent you pay for living on this wonderful planet.
3

ETL FUNCTIONS
The ETL process consists of
data extraction from source systems
data transformation which includes data cleaning and
data loading in the ODS or the data warehouse
Data cleaning deals with detecting & removing errors/inconsistencies from the data, in particular the
data that is sourced from a variety of computer systems.
Building an integrated database from a number of source-systems may involve solving some of the
following problems:
Instance Identity Problem
The same customer may be represented slightly differently in different source-systems.
There is a possibility of mismatching between the different systems that needs to be
identified & corrected.
Checking for homonyms & synonyms is necessary.
Achieving very high consistency in names & addresses requires a huge amount of
resources.
Data Errors
Following are different types of data errors
data may have some missing attribute values
there may be duplicate records
there may be wrong aggregations
there may be inconsistent use of nulls, spaces and empty values
some attribute values may be inconsistent(i.e. outside their domain)
there may be non-unique identifiers
Record Linkage Problem
This deals with problem of linking information from different databases that relates to
the same customer.
Record linkage can involve a large number of record comparisons to ensure linkages
that have a high level of accuracy.
Semantic Integration Problem
This deals with integration of information found in heterogeneous OLTP & legacy
sources.
Some of the sources may be relational.
Some may be even in text documents
Some data may be character strings while others may be integers.
Data Integrity Problem
This deals with issues like
referential integrity
null values
domain of values
Data cleaning should be based on the following 5 steps:
1) Parsing
This involves
identifying various components of the source data-files and
then establishing relationships between those and the fields in the target files.
2) Correcting
Correcting the identified components is usually based on a variety of sophisticated
techniques including mathematical algorithms.
3) Standardizing
Business rules of enterprise may be used to transform the data to standard form.
4) Matching
Much of the data extracted from a number of source-systems is likely to be related.
Such data needs to be matched.
5) Consolidating
All corrected, standardized and matched data can now be consolidated to build a single
version of the enterprise-data.
Happiness is our birthright.

4

DATA WAREHOUSE
This is an integrated subject-oriented & time-variant repository of information in support of
management's decision making process.
In other words, it is a process of integrating enterprise-wide data, originating from a variety of
sources, into a single repository.
ODS is a current-valued data store
while a DW is a time-variant repository of data (Table: 7.6).
Benefits of implementing a data warehouse:
1) To provide a single version of truth about enterprise information
2) To speed up adhoc reports and queries that involves aggregations across many attributes
3) To provide a system in which managers who do not have a strong technical background
are able to run complex queries
4) To provide a database that stores relatively clean data
5) To provide a DB that stores historical data that may have been deleted from OLTP systems
Q: Why a data warehouse must be separate from OLTP systems?
Ans: Complex queries involving number of aggregations are likely to degrade performance of systems.
Smaller data warehouse is called data marts (Figure: 7.3).
A data-mart stores information for a limited number of subject-areas.
The data-mart approach is attractive since beginning with a single data-mart is relatively inexpensive
and easier to implement.
Fear is the number one factor that causes people to live small, inauthentic lives.
5
ODS & DW ARCHITECTURE

The architecture involves
extracting information from source systems by using an ETL process and
then storing the information in a staging area (Figure: 7.4).
Staging area is a temporary database, separate from source systems and ODS where in the data is
copied from source systems
then transformed and
finally copied to the ODS.
The daily changes also come to the staging area.
Another ETL process is used to transform information from the staging area to populate the ODS.
The ODS could then be used for producing a variety of reports for management.
The ODS is then used for supplying information via another ETL process to the data warehouse which
in turn feeds a number of data marts that generate the reports required by management.
Focus on being great at what u do. Be truly outstanding in every element of ur professional and ur personal life.
6

DATA WAREHOUSE DESIGN
A dimension (essentially an attribute) is an ordinate within a multidimensional structure consisting of
a list of ordered values (called members).
A member is a distinct value (or name or identifier) for the dimension.
Values that depend on the dimensions are called measures.
The fact table is a collection of related data items, consisting of values of dimensions of interest and
the value of measure (Figure: 7.8).
A data warehouse model consists of
a central fact table &
a set of surrounding dimensions tables on which the facts depend. Such a model is called a
star scheme (Figure: 7.6).
The characteristic of a star scheme is that all the dimensions directly link to the fact table.
Star schemas may be refined into snowflake schemas if we wish to provide support for dimension
hierarchies by allowing dimension tables to have subtables to represent the hierarchies (Fig: 7.7).
.
If you can concentrate only on the present, youll be a happy man.

7
When you know better, you can do better.

8

DW IMPLEMENTATION STEPS
1) Requirements Analysis & Capacity Planning
The first step involves
defining enterprise needs
defining architecture
carrying out capacity planning and
selecting the hardware and software tools
This step also involves consulting
with senior management &
with the various stakeholders
2) Hardware Integration
Both hardware and software need to be put together by integrating
servers
storage devices &
client software tools
3) Modeling
This involves designing the warehouse schema and views. This may involve using a
modeling tool if the data warehouse is complex.
4) Physical Modeling
This involves designing
data warehouse organization
data placement
data partitioning
deciding on access methods and indexing
5) Sources
This involves identifying and connecting the sources using gateways.
6) ETL
This involves
identifying a suitable ETL tool vendor and
purchasing and implementing the tool
This may include customizing the tool to suit the needs of the enterprise.
7) Populate DW
This involves testing the required ETL tools, perhaps using a staging area. Then, ETL
tools may be used in populating the warehouse.
8) User Applications
This involves designing and implementing applications required by the end users.
9) Roll-out the DW and Applications
DW METADATA
Metadata is data about data or documentation about the data that is needed by the users.
This is not the actual data warehouse, but answers the "who, what, where, when, why and how"
questions about the data warehouse.
This describes the characteristics of a resource.
Metadata may be classified into two groups: back room metadata and front room metadata.
1) Much important information is included in the back room metadata that is process related,
for example, the ETL processes.
Furthermore, this could include
data extraction, cleaning and loading specifications
dates on which the ETL process was carried out and
a list of data that were loaded
2) The front room metadata is more descriptive and could include information needed by the
users, for example,
user security privileges
various statistics about usage
Stop being a prisoner of your past & become architect of your future.
9

DW IMPLEMENTATION GUIDELINES
Build Incrementally
Firstly, a data mart may be built with one particular project in mind.
Then, a number of other sections of enterprise may be implemented in a similar
manner.
An enterprise data warehouse can then be implemented in an iterative manner,
allowing all data marts to extract information from the data warehouse.
Need a Champion
The project must have a champion who is willing to carry out considerable research
into expected costs and benefits of the project.
The projects require inputs from many units in an enterprise and therefore need to be
driven by someone who is capable of interacting with people in the enterprise.
Senior Management Support
Given the resource intensive nature of such projects and the time they can take to
implement, the project calls for a sustained commitment from senior management.
Ensure Quality
Only data that has been cleaned and only data that are of a quality should be loaded in
the data warehouse.
Corporate Strategy
The project must fit with corporate strategy and business objectives.
Business Plan
Financial costs, expected benefits & a project plan must be clearly outlined and
understood by all stakeholders.
Training
The users must be trained
to use the warehouse and
to understand its capabilities
Adaptability
Project should build in adaptability so that changes may be made to DW if & when
required.
Joint mManagement
The project must be managed by both IT and business professionals in the
enterprise.
EXERCISES
1) Explain ODS along with its structure. (6)
2) Explain ETL. List out the issues that need to be resolved for successful
implementation of an ETL system. (6)
3) What is data cleaning? Explain the 5 steps in data cleaning. (4)
4) Explain the problems that need to be solved for building an integrated database from
a number of source systems. (6)
5) Write a short note on
i) Data warehouse (4)
ii) ZLE (4)
6) Compare the following:
i) Lazy query vs. eager query processing (4)
ii) OLTP vs. DW (6)
iii) ODS vs. DW (6)
7) Explain ODS & DW architecture along with its structure. (4)
8) What are star & snowflake schemas? (2)
9) Explain the steps for DW implementation. (6)
10) Explain the guidelines for DW implementation. (6)
11) Write a short note on DW metadata. (4)
No matter how insignificant the thing you have to do, do it as well as you can. For it will be by those small things
that you shall be judged.
10
UNIT 2: ONLINE ANALYTICAL PROCESSING (OLAP)

INTRODUCTION
A dimension is an attribute within a multidimensional model consisting of a list of values (called
members).
A fact is defined by a combination of dimension values for which a non-null value exists.
The non-null values of facts are the numerical values stored in each data cube cell. They are called
measures.
A data cube computes aggregates over all subsets of dimensions specified in the cube.
OLAP
OLAP stands for Online Transaction Processing Systems.
This is primarily a software-technology concerned with fast analysis of enterprise-information.
In other words, OLAP is the dynamic enterprise analysis required to create, manipulate, animate &
synthesize information from exegetical, contemplative and formulaic data analysis models.
Business-Intelligence(BI) is used to mean both data-warehousing and OLAP.
In other words, BI is defined as a user-centered process of
exploring data, data-relationships and trends
thereby helping to improve overall decision-making.
MOTIVATIONS FOR USING OLAP

1) Understanding and Improving Sales
For an enterprise that has many products and many channels for selling the products,
OLAP can assist in finding the most popular products & the most popular channels.
In some cases, it may be possible to find the most profitable customers.
Analysis of business-data can assist in improving the enterprise-business.
2) Understanding and Reducing Costs of doing Business
OLAP can assist in
analyzing the costs associated with sales &
controlling the costs as much as possible without affecting sales
In some cases, it may also be possible to identify expenditures that produce a high ROI
(return on investment).
What we do have enormous control over is the way we respond to what life sends our way.
11

CHARACTERISTICS OF OLAP SYSTEMS
1) Users
OLTP systems are designed for office-workers (100-1000 users).
Whereas, OLAP systems are designed for decision-makers(few business-users).
2) Functions
OLTP systems are mission-critical. They support an enterprise's day-to-day operations. They
are mostly performance-driven.
Whereas, OLAP systems are management-critical. They support an enterprise's decisionfunctions using analytical-investigations.
3) Nature
OLTP systems are designed to process one record at a time, for ex, a record related to the
customer.
Whereas, OLAP systems involve queries that deal with many records at a time and
provide aggregate data to a manager.
4) Design
OLTP systems are designed to be application-oriented.
Whereas, OLAP systems are designed to be subject-oriented.
OLTP systems view the enterprise-data as a collection of tables (based on ER model).
Whereas, OLAP systems view enterprise-information as multidimensional model.
5) Data
OLTP systems normally deal only with the current-status of information. The old information
may have been archived and may not be accessible online.
Whereas, OLAP systems require historical-data over several years.
6) Kind of use
OLTP systems are used for read & write operations.
Whereas, OLAP systems normally do not update the data.
Actions have consequences and to reap the harvest you dream of, you must sow the seeds.
12

FASMI CHARACTERISTICS OF OLAP SYSTEMS
Fast
Most queries should be answered very quickly, perhaps within seconds.
The performance of the system has to be like that of a search-engine.
The data-structures must be efficient.
The hardware must be powerful enough for
amount of data &
number of users
One approach can be
pre-compute the most commonly queried aggregates and
compute the remaining aggregates on-the-fly
Analytic
The system must provide rich analytic-functionality.
Most queries should be answered without any programming.
The system should be able to cope with any relevant queries for application & user.
Shared
The system is
likely to be accessed only by few business-analysts and
may be used by thousands of users
Being a shared system, the OLAP software should provide adequate security for
confidentiality & integrity.
Concurrency-control is obviously required if users are writing or updating data in the
database.
Multidimensional
This is the basic requirement.
OLAP software must provide a multidimensional conceptual view of the data.
A dimension often has hierarchies that show parent/child relationships between the
members of dimensions. The multidimensional structure should allow such hierarchies.
Information
The system should be able to handle a large amount of input-data.
The capacity of system to handle information and its integration with the data
warehouse may be critical.
CODD'S OLAP CHARACTERISTICS
Multidimensional Conceptual View
This is the central characteristics.
By requiring a multidimensional view, it is possible to carry out operations like slice
and dice.
Accessibility (OLAP as a Mediator)
The OLAP software should be sitting between data-sources (e.g. data warehouse) and
an OLAP front-end.
Batch Extraction vs. Interpretive
The system should provide multidimensional data staging plus partial pre-calculation
of aggregates in large multidimensional databases.
Multi-user Support
Being a shared system, the OLAP software should provide many normal database
operations including retrieval, update, concurrency-control, integrity and security.
Storing results of OLAP
OLAP results data should be kept separate from source-data.
Read-write OLAP applications should not be implemented directly on live transactiondata if OLTP source systems are supplying information to the OLAP system directly.
Extraction of Missing Values
The system should distinguish missing-values from zero-values. If a distinction is not
made, the aggregates are likely to be computed incorrectly.
Uniform Reporting Performance
Increasing the number of dimensions (or database-size) should not significantly
degrade the reporting performance of the system.
Commitment and accountability are words that must flood the core of your being.
13

Treatment of Missing Values
The system should ignore all missing-values regardless of their source.
Generic Dimensionality
The system should treat each dimension as equivalent in both its structure and
operational capabilities.
Unlimited Dimensions and Aggregation Levels
The system should allow unlimited dimensions and aggregation levels.
In practice, number of dimensions is rarely more than 10 and
number of hierarchies rarely more than 6.
MULTIDIMENSIONAL VIEW AND DATA CUBE
The multidimensional view of data is in some ways a natural view of any enterprise for managers
The triangle shows that as we go higher in the triangle hierarchy the managers need for detailed
information declines (Figure 8.1).
The multidimensional view of data by using an example of a simple OLTP database consists of the
three tables:
student(Student_id, Student_name, Country, DOB, Address)
enrolment(Student_id, Degree_id, SSemester)
degree(Degree_id, Degree_name, Degree_length, Fee, Department)
Learning should be fun, memorable and moving.

14
It is clear that the information given in Tables 8.3, 8.4 and 8.5, although suitable for a student
enrolment OLTP system, is not suitable for efficient management decision making.
The managers do not need information about the individual students, the degree they are enrolled in,
and the semester they joined the university.
What the managers needs the trends in student numbers in different degree programs and from
different countries.
We first consider only two dimensions. Let us say we are primarily interested in finding out how
many students from each country came to do a particular degree (Table: 8.6). Therefore we may
visualize the data as two-dimensional, i.e.,
Country x Degree
Using this two-dimensional view we are able to find the number of students joining any degree from
any country (only for semester 2000-01). Other queries that we are quickly able to answer are:
How many students started BIT in 2000-01?
How many students joined from Singapore in 2000-01?
Experience is always the best teacher.

15
Tables 8.6, 8.7 and 8.8 together now form a three-dimensional cube. The table 8.7 provides totals
for the two semesters and we are able to drill-down to find numbers in individual semesters.
Putting the three tables together gives a cube of 8 x 6 x 3 (= 144) cells including the totals along
every dimension.
A cube could be represented by:
Country x Degree x Semester
Figure 8.2: cube formed by tables 8.6, 8.7 & 8.8
In the three-dimensional cube, the following eight types of aggregations or queries are
possible:
1. null (e.g. how many students are there? Only 1 possible query)
2. degrees (e.g. how many students are doing BSc? 5 possible queries if we assume 5 different
degrees)
3. semester (e.g. how many students entered in semester 2000-01? 2 possible queries if we
only have data about 2 semesters)
4. country (e.g. how many students are from the USA? 7 possible queries if there are 7
countries)
5. degrees, semester (e.g. how many students entered in 2000-01 to enroll in BCom? With 5
degrees and 2 different semesters 10 queries)
6. semester, country (e.g. how many students from the UK entered in 2000-01? With 7
countries and 2 different semesters 14 queries)
7. degrees, country (e.g. how many students from Singapore are enrolled in BCom? 7*5=35
queries)
8. all (e.g. how many students from Malaysia entered in 2000-01 to enroll in BCom? 7*5*2=70
queries)
All the cell in the cube represents measures or aggregations.
2n types of aggregations are possible for n dimensions.
Devote yourself to displaying a standard of excellence at work far higher than anyone would ever expect from you.
16

DATA CUBE IMPLEMENTATION
Pre-compute and store all
Millions of aggregates need to be computed and stored.
This is the best solution as far as query response-time is concerned.
This solution is impractical for a large data-cube,
since resources required to compute & store the aggregates will be prohibitively
large
Indexing large amounts of data is also expensive.
Pre-compute(and store) none
The aggregates are computed on-the-fly using the raw data whenever a query is
posed.
This approach does not require additional space for storing the cube.
The query response-time is likely to be very poor for large data-cubes.
Pre-compute and store some
We pre-compute and store the most frequently queried aggregates and compute
others as the need arises.
We may also be able to derive some of the remaining aggregates using the aggregates
that have already been computed (Figure 8.3). .
The more aggregates we are able to pre-compute, the better the query performance
Data-cube products use different techniques for pre-computing aggregates and storing
them. They are generally based on one of two implementation models.
i) ROLAP (relational OLAP)
ii) MOLAP (multidimensional OLAP)
Asking the right question is often 90% of finding the right answer.
17

ROLAP
This uses relational or extended-relational DBMS to store and manage warehouse data.
This may be considered a bottom-up approach to OLAP.
This is typically based on using a data warehouse that has been designed using a star scheme.
The data warehouse provides the multidimensional capabilities by representing data in fact-table and
dimension-tables.
The fact-table contains one column for each dimension and one column for each measure and every
row of the table provides one fact.
A fact then is represented as with the last column as 30. An OLAP tool is then provided to manipulate
the data in these data warehouse tables.
This tool essentially groups the fact-table to find aggregates and uses some of the aggregates
already computed to find new aggregates.
Advantages:
This is more easily used with existing relational DBMS and
The data can be stored efficiently using tables.
Greater scalability
Disadvantage:
Poor query performance
Some products in this category are Oracle OLAP mode and OLAP Discoverer.
MOLAP
This is based on using a multidimensional DBMS rather than a data-warehouse to store and access
data.
This may be considered as a top-down approach to OLAP.
This does not have a standard approach to storing and maintaining their data.
This often uses special-purpose file systems or indexes that store pre-computation of all aggregations
in the cube.
Advantages:
Implementation is usually exceptionally efficient
Easier to use and therefore may be more suitable for inexperienced users
Fast indexing to pre-computed summarized-data
Disadvantages:
More expensive than ROLAP
Data is not always current
Difficult to scale a MOLAP system for very large OLAP problems
Storage utilization may be low if the data-set is sparse
Some MOLAP products are Hyperion Essbase and Applix iTM1.
Life is short and you dont know when it's going to end.
18

DATA CUBE OPERATIONS
A number of operations may be applied to data cubes. The common ones are:
roll-up
drill-down
pivot or rotate
slice & dice
ROLL-UP
This is like zooming out on the data cube.
This is required when the user needs further abstraction or less detail.
Initially the concept hierarchy was "street < city < province < country".
On rolling up the data is aggregated by ascending the location hierarchy from the level of city to level
of country.
The data is grouped into cities rather than countries.
DRILL DOWN
This is like zooming in on the data and is therefore the reverse of roll-up.
This is an appropriate operation when the user needs further details or when the user wants to
partition more finely or wants to focus on some particular values of certain dimensions.
This adds more details to the data.
Initially the concept hierarchy was "day < month < quarter < year."
On drill-up the time dimension is descended from the level quarter to the level of month.
When drill-down operation is performed then one or more dimensions from the data cube are added.
Life's a game. Dont take it too seriously. Have fun. Dance. Laugh. Maintain a healthy dose of perspective.
19
PIVOT OR ROTATE
This is used when the user wishes to re-orient the view of the data cube.
This may involve
swapping the rows and columns, or
moving one of the row dimensions into the column dimension
In this, the item and location axes in 2-D slice are rotated.
There are no great acts, only small acts done with great love.
20

SLICE & DICE
These are operations for browsing the data in the cube. The terms refer to the ability to look at
information from different viewpoints.
A slice is a subset of cube corresponding to a single value for 1 or more members of dimensions.
The slice operation is performed for the dimension time using the criterion time ="Q1".
The dice operation is similar to slice but dicing does not involve reducing number of dimensions.
A dice is obtained by performing a selection on two or more dimensions.
The dice operation on cube based on the following selection criteria that involve three dimensions.
(location = "Toronto" or "Vancouver")
(time = "Q1" or "Q2")
(item =" Mobile" or "Modem").
Life shrinks or expands in proportion to one's courage.

21

GUIDELINES FOR OLAP IMPLEMENTATION
Vision
The OLAP team must, in consultation with the users, develop a clear vision for the
OLAP system.
This vision (including the business-objectives) should be clearly defined, understood
and shared by the stakeholders.
Senior Management Support
The project should be fully supported by the senior-managers.
Since a data warehouse may have been developed already, this should not be
difficult.
Selecting an OLAP Tool
The team should familiarize themselves with the ROLAP and MOLAP tools available in
the market.
Since tools are quite different, careful planning may be required in selecting a tool
that is appropriate for the enterprise.
Corporate Strategy
The OLAP-strategy should fit with the enterprise-strategy and business-objectives. A
good fit will result in the OLAP tools being used more widely.
Focus on the Users
The project should be focused on the users.
Users should, in consultation with the technical professionals, decide what tasks will
be done first and what will be done later.
A good GUI should be provided to non-technical users.
The project can only be successful with the full support of the users.
Joint Management
The project must be managed by both the IT and business professionals.
Many other people should be involved in supplying ideas.
Review and Adapt
Regular reviews of the project may be required to ensure that the project is meeting
the current needs of the enterprise.
OLAP SOFTWARE
SQL Server 2000 Analysis Service from Microsoft. SQL Server 2000 analysis services is the OLAP
services component in SQL Server 7.0.
BI2M(Business Intelligence to Marketing and Management) from B&M Service has 3 modules, one
of which is for OLAP. The OLAP module allows database exploring including slice and dice, roll-up,
drill-down and displays results as 2D Charts, 3D charts and tables.
Business Objects OLAP Intelligence from BusinessObjects allows access to OLAP servers from
Microsoft, Hyperion, IBM and SAP. BusinessObjects also has widely used Crystal Analysis and
Reports.
Usual operations like slice and dice, and drill directly on multidimensional sources are possible.
ContourCube from Contour Components is an OLAP product that enables users to slice and dice,
roll up, drill down and pivot efficiently.
DB2 Cube Views from IBM includes features and functions for managing and deploying
multidimensional data.
Express and the Oracle OLAP option. Express is a multidimensional database and application
development environment for building OLAP applications.
EXERCISES
1) What is OLAP? Explain the motivations for using OLAP. (4)
2) Compare the following:
i) OLTP vs. OLAP (6)
ii) ROLAP vs. MOLAP (4)
3) Explain the FASMI characteristics of OLAP systems. (6)
4) Explain the Codd's OLAP characteristics. (6)
5) Explain the 3 methods for data cube implementation. (6)
6) Explain various operations on data cube. (6)
7) Explain the guidelines for OLAP implementation. (6)
Aim for the top. There is plenty of room there. There are so few at the top it is almost lonely there.
22
UNIT 3: DATA MINING

WHAT IS DATA MINING?
Data Mining is the process of automatically discovering useful information in large datarepositories.
DM techniques
can be used to search large DB to find useful patterns that might otherwise remain unknown
provide capabilities to predict the outcome of future observations
Why do we need Data Mining?
Conventional database systems provide users with query & reporting tools.
To some extent the query & reporting tools can assist in answering questions like, where did the
largest number of students come from last year?
But these tools cannot provide any intelligence about why it happened.
Taking an Example of University Database System
The OLTP system will quickly be able to answer the query like how many students are enrolled in
university
The OLAP system using data warehouse will be able to show the trends in students enrollments
(ex: how many students are preferring BCA),
Data mining will be able to answer where the university should market.
DATA MINING AND KNOWLEDGE DISCOVERY
Data Mining is an integral part of KDD (Knowledge Discovery in Databases).
KDD is the overall process of converting raw data into useful information (Figure: 1.1).
The input-data is stored in various formats such as flat files, spread sheet or relational tables.
Purpose of preprocessing: to transform the raw input-data into an appropriate format for
subsequent analysis.
The steps involved in data-preprocessing include
combine data from multiple sources
clean data to remove noise & duplicate observations, and
select records & features that are relevant to the DM task at hand
Data-preprocessing is perhaps the most time-consuming step in the overall knowledge discovery
process.
Closing the loop" refers to the process of integrating DM results into decision support systems.
Such integration requires a postprocessing step. This step ensures that only valid and useful
results are incorporated into the decision support system.
An example of postprocessing is visualization.
Visualization can be used to explore data and DM results from a variety of viewpoints.
Statistical measures can also be applied during postprocessing to eliminate bogus DM results.
Difficulties increase the nearer we approach the goal.

23

MOTIVATING CHALLENGES
Scalability
Nowadays, data-sets with sizes of terabytes or even petabytes are becoming common.
DM algorithms must be scalable in order to handle these massive data sets.
Scalability may also require the implementation of novel data structures to access individual
records in an efficient manner.
Scalability can also be improved by developing parallel & distributed algorithms.
High Dimensionality
Traditional data-analysis technique can only deal with low dimensional data.
Nowadays, data-sets with hundreds or thousands of attributes are becoming common.
Data-sets with temporal or spatial components also tend to have high dimensionality.
The computational complexity increases rapidly as the dimensionality increases.
Heterogeneous and Complex Data
Traditional analysis methods can deal with homogeneous type of attributes.
Recent years have also seen the emergence of more complex data-objects.
DM techniques for complex objects should take into consideration relationships in the data, such as
temporal & spatial autocorrelation
parent-child relationships between the elements in semi-structured text & XML documents
Data Ownership & Distribution
Sometimes, the data is geographically distributed among resources belonging to multiple entities.
Key challenges include:
1) How to reduce amount of communication needed to perform the distributed computation
2) How to effectively consolidate the DM results obtained from multiple sources &
3) How to address data-security issues
Non Traditional Analysis
The traditional statistical approach is based on a hypothesized and test paradigm.
In other words, a hypothesis is proposed, an experiment is designed to gather the data, and then
the data is analyzed with respect to hypothesis.
Current data analysis tasks often require the generation and evaluation of thousands of
hypotheses, and consequently, the development of some DM techniques has been motivated by the
desire to automate the process of hypothesis generation and evaluation.
THE ORIGIN OF DATA MINING
Data mining draws upon ideas from
Sampling, estimation, and hypothesis test from statistics
Search algorithms, modeling techniques machine learning, learning theories from AI
pattern recognition, statistics database systems
Traditional techniques may be unsuitable due to
Enormity of data
High dimensionality of data
Heterogeneous nature of data
Data mining also had been quickly to adopt ideas from other areas including
Optimization
Evolutionary computing
Signal processing
Information theory
Database systems are needed to provide supports for efficient storage, indexing, query processing.
The parallel computing and distribute technology are two major data addressing issues in data
mining to increase the performance.
People with goals succeed because they know where they're going.
24

DATA MINING TASKS
DM tasks are generally divided into 2 major categories.
Predictive Tasks
The objective is to predict the value of a particular attribute based on the values of other
attributes.
The attribute to be predicted is commonly known as the target or dependent variable,
while the attributes used for making the predication are known as the explanatory or
independent variables.
Descriptive Tasks
The objective is to derive patterns (correlations, trends, clusters, trajectories and
anomalies) that summarize the relationships in data.
Descriptive DM tasks are often exploratory in nature and frequently require postprocessing
techniques to validate and explain the results.
Four of the Core Data Mining Tasks
1) Predictive Modeling
This refers to the task of building a model for the target variable as a function of the
explanatory variable.
The goal is to learn a model that minimizes the error between the predicted and true values
of the target variable.
There are 2 types of predictive modeling tasks:
i) Classification: used for discrete target variables
Ex: Web user will make purchase at an online bookstore is a classification task,
because the target variable is binary valued.
ii) Regression: used for continuous target variables.
Ex: forecasting the future price of a stock is regression task because price is a
continuous values attribute
2) Cluster Analysis
This seeks to find groups of closely related observations so that observations that belong to
the same cluster are more similar to each other than observations that belong to other
clusters.
Clustering has been used
to group sets of related customers
to find areas of the ocean that have a significant impact on the Earth's climate
We all have two choices; We can make a living or we can design a life.
25

3) Association Analysis
This is used to discover patterns that describe strongly associated features in the data.
The goal is to extract the most interesting patterns in an efficient manner.
Useful applications include
finding groups of genes that have related functionality or
identifying web pages that are accessed together
Ex: market based analysis
We may discover the rule that {diapers} -> {Milk}, which suggests that customers who buy
diapers also tend to buy milk.
4) Anomaly Detection
This is the task of identifying observations whose characteristics are significantly different
from the rest of the data. Such observations are known as anomalies or outliers.
The goal is
to discover the real anomalies and
to avoid falsely labeling normal objects as anomalous.
Applications include the detection of fraud, network intrusions, and unusual patterns of
disease.
Example 1.4 (Credit Card Fraud Detection).
A credit card company records the transactions made by every credit card holder,
along with personal information such as credit limit, age, annual income, and address.
Since the number of fraudulent cases is relatively small compared to the number of
legitimate transactions, anomaly detection techniques can be applied to build a profile of
legitimate transactions for the users.
When a new transaction arrives, it is compared against the profile of the user.
If the characteristics of the transaction are very different from the previously created
profile, then the transaction is flagged as potentially fraudulent
EXERCISES
1. What is data mining? Explain Data Mining and Knowledge Discovery? (10)
2. What are different challenges that motivated the development of DM? (10)
3. Explain Origins of data mining (5)
4. Discuss the tasks of data mining with suitable examples. (10)
5. Explain Anomaly Detection .Give an Example? (5)
6. Explain Descriptive tasks in detail? (10)
7. Explain Predictive tasks in detail by example? (10)
Success is not measured in achievement of goals, but in the stress and strain of meeting those goal.
26
UNIT 3: DATA MINING (CONT.)

WHAT IS A DATA OBJECT?
A data-set refers to a collection of data-objects and their attributes.
Other names for a data-object are record, transaction, vector, event, entity, sample or
observation.
Data-objects are described by a number of attributes such as
mass of a physical object or
time at which an event occurred.
Other names for an attribute are dimension, variable, field, feature or characteristics.
WHAT IS AN ATTRIBUTE?
An attribute is a characteristic of an object that may vary, either
from one object to another or
from one time to another.
For example, eye color varies from person to person.
Eye color is a symbolic attribute with a small no. of possible values {brown, black, blue, green}.
Example 2.2 (Student Information).
Often, a data-set is a file, in which the objects are records(or rows) in the file and each field
(or column) corresponds to an attribute.
For example, Table 2.1 shows a data-set that consists of student information.
Each row corresponds to a student and each column is an attribute that describes some
aspect of a student, such as grade point average(GPA) or identification number(ID).
PROPERTIES OF ATTRIBUTE VALUES

The type of an attribute depends on which of the following properties it possesses:
1) Distinctness: =
2) Order: < >
3) Addition: + 4) Multiplication: * /
Nominal attribute: Uses only distinctness.
Examples: ID numbers, eye color, pin codes
Ordinal attribute: Uses distinctness & order.
Examples: Grades in {SC, FC, FCD}
Shirt sizes in {S, M, L, XL}
Interval attribute: Uses distinctness, order & addition
Examples: calendar dates, temperatures in Celsius or Fahrenheit.
Ratio attribute: Uses all 4 properties
Examples: temperature in Kelvin, length, time, counts
When the promise is clear, the price gets easy.

27

DIFFERENT TYPES OF ATTRIBUTES
DESCRIBING ATTRIBUTES BY THE NUMBER OF VALUES

1) Discrete
Has only a finite or countably infinite set of values.
Examples: pin codes, ID Numbers, or the set of words in a collection of documents.
Often represented as integer variables.
Binary attributes are a special case of discrete attributes and assume only 2 values.
E.g. true/false, yes/no, male/female or 0/1
2) Continuous
Has real numbers as attribute values.
Examples: temperature, height, or weight.
Often represented as floating-point variables.
ASYMMETRIC ATTRIBUTES
Binary attributes where only non-zero values are important are called asymmetric binary
attributes.
Consider a data-set where each object is a student and each attribute records whether or not a
student took a particular course at a university.
For a specific student, an attribute has a value of 1 if the student took the course associated with
that attribute and a value of 0 otherwise.
Because students take only a small fraction of all available courses, most of the values in such a
data-set would be 0.
Therefore, it is more meaningful and more efficient to focus on the non-zero values.
This type of attribute is particularly important for association analysis.
Live like tomorrow will not come.
28

TYPES OF DATA SETS
1) Record data
Transaction (or Market based data)
Data matrix
Document data or Sparse data matrix
2) Graph data
Data with relationship among objects (World Wide Web)
Data with objects that are Graphs (Molecular Structures)
3) Ordered data
Sequential data (Temporal data)
Sequence data
Time series data
Spatial data
GENERAL CHARACTERISTICS OF DATA SETS
Following 3 characteristics apply to many data-sets:
1) Dimensionality
Dimensionality of a data-set is no. of attributes that the objects in the data-set
possess.
Data with a small number of dimensions tends to be qualitatively different than
moderate or high-dimensional data.
The difficulties associated with analyzing high-dimensional data are sometimes
referred to as the curse of dimensionality.
Because of this, an important motivation in preprocessing data is dimensionality
reduction.
2) Sparsity
For some data-sets with asymmetric feature, most attribute of an object have values
of 0.
In practical terms, sparsity is an advantage because usually only the non-zero
values need to be stored & manipulated.
This results in significant savings with respect to computation-time and storage.
Some DM algorithms work well only for sparse data.
3) Resolution
This is frequently possible to obtain data at different levels of resolution, and often
the properties of the data are different at different resolutions.
Ex: the surface of the earth seems very uneven at a resolution of few meters, but is
relatively smooth at a resolution of tens of kilometers.
The patterns in the data also depend on the level of resolution.
If the resolution is too fine, a pattern may not be visible or may be buried in noise.
If the resolution is too coarse, the pattern may disappear.
Aim for the stars and maybe you'll reach the sky
29

RECORD DATA
Data-set is a collection of records.
Each record consists of a fixed set of attributes.
Every record has the same set of attributes.
There is no explicit relationship among records or attributes.
The data is usually stored either
in flat files or
in relational databases
TYPES OF RECORD DATA

1) Transaction (Market Basket Data)
Each transaction consists of a set of items.
Consider a grocery store.
The set of products purchased by a customer represents a transaction
while the individual products represent items.
This type of data is called market basket data because
the items in each transaction are the products in a person's "market basket."
Data can also be viewed as a set of records whose fields are asymmetric attributes.
2) Data Matrix
An m*n matrix, where there are m rows, one for each object, &
n columns, one for each attribute.
This matrix is called a data-matrix.
Since data-matrix consists of numeric attributes, standard matrix operation can be
applied to manipulate the data.
3) Sparse Data Matrix
This is a special case of a data-matrix.
The attributes are of the same type and are asymmetric i.e. only non-zero values
are important.
Document Data
A document can be represented as a vector,
where each term is a attribute of the vector and
value of each attribute is the no. of times corresponding term
occurs in the document.
Wise people remind themselves that every day could be their last.
30

GRAPH BASED DATA
Sometimes, a graph can be a convenient and powerful representation for data.
We consider 2 specific cases:
1) Data with Relationships among Objects
The relationships among objects frequently convey important information.
In particular, the data-objects are mapped to nodes of the graph,
while relationships among objects are captured by link properties such as
direction & weight.
For ex, in web, the links to & from each page provide a great deal of information about
the relevance of a web-page to a query, and thus, must also be taken into consideration.
2) Data with Objects that are Graphs
If the objects contain sub-objects that have relationships, then such objects are
frequently represented as graphs.
For ex, the structure of chemical compounds can be represented by a graph,
where nodes are atoms and
links between nodes are chemical bonds.
No one knows what he can do until he tries.

31

ORDERED DATA
Sequential Data (Temporal Data)
This can be thought of as an extension of record-data,
where each record has a time associated with it.
A time can also be associated with each attribute.
For example, each record could be the purchase history of a customer, with a listing of
items purchased at different times.
Using this information, it is possible to find patterns such as "people who buy DVD
players tend to buy DVDs in the period immediately following the purchase."
Sequence Data
This consists of a data-set that is a sequence of individual entities, such as a sequence
of words or letters.
This is quite similar to sequential data, except that there are no time stamps; instead,
there are positions in an ordered sequence.
For example, the genetic information of plants and animals can be represented in the
form of sequences of nucleotides that are known as genes.
Time Series Data
This is a special type of sequential data in which a series of measurements are taken
over time.
For example, a financial data-set might contain objects that are time series of the daily
prices of various stocks.
An important aspect of temporal-data is temporal-autocorrelation i.e. if two
measurements are close in time, then the values of those measurements are often very
similar.
Spatial Data
Some objects have spatial attributes, such as positions or areas.
An example is weather-data (temperature, pressure) that is collected for a variety of
geographical location.
An important aspect of spatial-data is spatial-autocorrelation i.e. objects that are
physically close tend to be similar in other ways as well.
Do not let the future be held hostage by the past.

32

DATA PREPROCESSING
Data preprocessing is a broad area and consists of a number of different strategies and techniques
that are interrelated in complex way.
Different data processing techniques are:
1. Aggregation
2. Sampling
3. Dimensionality reduction
4. Feature subset selection
5. Feature creation
6. Discretization and binarization
7. Variable transformation
AGGREGATION
This refers to combining 2 or more attributes (or objects) into a single attribute (or object).
For example, merging daily sales figures to obtain monthly sales figures
Motivations for aggregation:
1) Data reduction: The smaller data-sets require
less memory
less processing time.
Because of aggregation, more expensive algorithm can be used.
2) Aggregation can act as a change of scale by providing a high-level view of the data instead
of a low-level view.
E.g. Cities aggregated into districts, states, countries, etc
3) The behavior of groups of objects is often more stable than that of individual objects.
Disadvantage: The potential loss of interesting details.
SAMPLING
This is a method used for selecting a subset of the data-objects to be analyzed.
This is often used for both
preliminary investigation of the data
final data analysis
Q: Why sampling?
Ans: Obtaining & processing the entire set of data of interest is too expensive or time consuming.
Sampling can reduce data-size to the point where better & more expensive algorithm can be used.
Key principle for effective sampling: Using a sample will work almost as well as using entire data-set,
if the sample is representative.
Sampling Methods
1) Simple Random Sampling
There is an equal probability of selecting any particular object.
There are 2 variations on random sampling:
i) Sampling without Replacement
As each object is selected, it is removed from the population.
ii) Sampling with Replacement
Objects are not removed from the population as they are selected for the sample.
The same object can be picked up more than once.
When the population consists of different types(or number) of objects, simple random
sampling can fail to adequately represent those types of objects that are less frequent.
2) Stratified Sampling
This starts with pre-specified groups of objects.
In the simplest version, equal numbers of objects are drawn from each group even though the
groups are of different sizes.
In another variation, the number of objects drawn from each group is proportional to the size
of that group.
3) Progressive Sampling
If proper sample-size is difficult to determine then progressive sampling can be used.
This method starts with a small sample, and
then increases the sample-size until a sample of sufficient size has been obtained.
This method requires a way to evaluate the sample to judge if it is large enough.
One of our biggest regrets on our deathbeds is that we were not reflective enough.
33

DIMENSIONALITY REDUCTION
Key benefit: many DM algorithms work better if the dimensionality is lower.
Purpose
May help to eliminate irrelevant features or reduce noise.
Can lead to a more understandable model (which can be easily visualized).
Reduce amount of time and memory required by DM algorithms.
Avoid curse of dimensionality.
The Curse of Dimensionality
Data-analysis becomes significantly harder as the dimensionality of the data increases.
For classification, this can mean that there are not enough data-objects to allow the creation
of a model that reliably assigns a class to all possible objects.
For clustering, the definitions of density and the distance between points (which are critical for
clustering) become less meaningful.
As a result, we get
reduced classification accuracy &
poor quality clusters.
FEATURE SUBSET SELECTION
Another way to reduce the dimensionality is to use only a subset of the features.
This might seem that such approach would lose information, this is not the case if redundant and
irrelevant features are present.
1) Redundant features duplicate much or all of the information contained in one or more other
attributes.
For example: purchase price of a product and the amount of sales tax paid.
2) Irrelevant features contain almost no useful information for the DM task at hand.
For example: students' ID numbers are irrelevant to the task of predicting students' grade
point averages.
Techniques for Feature Selection
1) Embedded approaches: Feature selection occurs naturally as part of DM algorithm.
Specifically, during the operation of the DM algorithm, the algorithm itself decides which
attributes to use and which to ignore.
2) Filter approaches: Features are selected before the DM algorithm is run.
3) Wrapper approaches: Use DM algorithm as a black box to find best subset of attributes.
An Architecture for Feature Subset Selection
The feature selection process is viewed as consisting of 4 parts:
1) A measure of evaluating a subset,
2) A search strategy that controls the generation of a new subset of features,
3) A stopping criterion and
4) A validation procedure.
Opportunity is always knocking. The problem is that most people have the self-doubt station in their head turned
up way too loud to
34

DISCRETIZATION AND BINARIZATION
Some DM algorithms (especially classification algorithms) require that the data be in the form of
categorical attributes.
Algorithms that find association patterns require that the data be in the form of binary attributes.
Transforming continuous attributes into a categorical attribute is called discretization.
And transforming continuous & discrete attributes into binary attributes is called as binarization.
Binarization
A simple technique to binarize a categorical attribute is the following: If there are m
categorical values, then uniquely assign each original value to an integer in interval [0,m-1].
Next, convert each of these m integers to a binary number.
Since n=[log2(m)] binary digits are required to represent these integers, represent these
binary numbers using n binary attributes.
Discretizction of continuous attributes

Discretization is typically applied to attributes that are used in classification or association
analysis.
In general, the best discretization depends on
the algorithm being used, as well as
the other attributes being considered
Transformation of a continuous attribute to a categorical attribute involves two subtasks:
deciding how many categories to have and
determining how to map the values of the continuous attribute to these categories
VARIABLE TRANSFORMATION
This refers to a transformation that is applied to all the values of a variable.
Ex: converting a floating point value to an absolute value.
Two types are:
1) Simple Functions
A simple mathematical function is applied to each value individually.
If x is a variable, then examples of transformations include ex, 1/x, log(x), sin(x).
2) Normalization (or Standardization)
The goal is to make an entire set of values have a particular property.
A traditional example is that of "standardizing a variable" in statistics.
If x is the mean of the attribute values and sx is their standard deviation, then the
transformation x'=(x- x )/sx creates a new variable that has a mean of 0 and a
standard deviation of 1.
Success is a journey, not a destination.

35

MEASURE OF SIMILARITY AND DISSIMILARITY
Similarity & dissimilarity are important because
they are used by a no. of DM techniques such as clustering, classification & anomaly detection.
Proximity is used to refer to either similarity or dissimilarity.
The similarity between 2 objects is a numerical measure of degree to which the 2 objects are alike.
Consequently, similarities are higher for pairs of objects that are more alike.
Similarities are usually non-negative and are often between 0(no similarity) and 1(complete
similarity).
The dissimilarity between 2 objects is a numerical measure of the degree to which the 2 objects are
different.
Dissimilarities are lower for more similar pairs of objects.
The term distance is used as a synonym for dissimilarity.
Dissimilarities sometimes fall in the interval [0,1] but is also common for them to range from 0 to
infinity.
Winners are those people who make a habit of doing the things losers are uncomfortable doing.
36

DISSIMILARITIES BETWEEN DATA OBJECTS
Distances
The Euclidean distance, d, between 2 points, x and y, in one-,two-,three- or higher-dimensional
space, is given by the following familiar formula:
where n=number of dimensions

xk and yk are respectively the kth attributes of x and y.
The Euclidean distance measure given in equation 2.1 is generalized by the Minkowski distance
metric given by
where r=parameter.
The following are the three most common examples of minkowski distance:
r=1. City block( Manhattan L1 norm) distance.
A common example is the Hamming distance, which is the number of bits that are different
between two objects that have only binary attributes ie between two binary vectors.
r=2. Euclidean distance (L2 norm)
r=. Supremum(L or Lmax norm) distance. This is the maximum difference between any
attribute of the objects. Distance is defined by
If d(x,y) is the distance between two points, x and y, then the following properties hold
1) Positivity
d(x,x)>=0 for all x and y
d(x,y)=0 only if x=y
2) Symmetry
d(x,y)=d(y,x) for all x and y.
3) Triangle inequality
d(x,z)<=d(x,y)+d(y,z) for all points x,y and z.
Measures that satisfy all three properties are known as metrics.
Let go of the past and go for the future. Go confidently in the direction of your dreams. Live the life you've imagined.
37

SIMILARITIES BETWEEN DATA OBJECTS
For similarities, the triangle inequality typically does not hold, but symmetry positivity typically do.
If s(x,y) is the similarity between points x and y, then the typical properties of similarities are the
following
1) s(x,y)=1 only if x=y
2) s(x,y)=s(y,x) for all x and y. (Symmetry)
For ex, cosine and Jaccard similarity.
EXAMPLES OF PROXIMITY MEASURES
Similarity measures between objects that contain only binary attributes are called similarity
coefficients.
Typically, they have values between 0 and 1.
A value of 1 indicates that the two objects are completely similar,
while a value of 0 indicates that the objects are not at all similar.
Let x and y be 2 objects that consist of n binary attributes.
Comparison of 2 objects, ie, 2 binary vectors, leads to the following four quantities (frequencies):
f00=the number of attributes where x is 0 and y is 0.
SIMPLE MATCHING COEFFICIENT
One commonly used similarity coefficient is the SMC, which is defined as
This measure counts both presences and absences equally.

JACCARD COEFFICIENT
Jaccard coefficient is frequently used to handle objects consisting of asymmetric binary attributes.
The jaccard coefficient is given by the following equation:
Courage means to continue seeking solutions to difficult problems, and to stay focused during stressful periods.
38

COSINE SIMILARITY
Documents are often represented as vectors, where each attribute represents the frequency with
which a particular term (or word) occurs in the document.
This is more complicated, since certain common words are ignored and various processing techniques
are used to account for
different forms of the same word
differing document lengths and
different word frequencies.
The cosine similarity is one of the most common measure of document similarity.
If x and y are two document vectors, then
As indicates by figure 2.16, cosine similarity really is a measure of the angle between x and y.
Thus, if the cosine similarity is 1,the angle between x and y is 0',and x and y are the same except for
magnitude(length).
If cosine similarity is 0, then the angle between x and y is 90' and they do not share any terms.
EXTENDED JACCARD COEFFICIENT(TANIMOTO COEFFICIENT)

This can be used for document data.
this coefficient is defined by following equation
ISSUES IN PROXIMITY CALCULATION

1) How to handle the case in which attributes have different scales and/or are correlated.
2) How to calculate proximity between objects that are composed of different types of attributes
e.g. quantitative and qualitative.
3) How to handle proximity calculation when attributes have different weights.
COMBINING SIMILARITIES FOR HETEROGENEOUS ATTRIBUTES
A general approach is needed when the attributes are of different types.
One straightforward approach is to compute the similarity between each attribute separately and
then combine these similarities using a method that results in a similarity between 0 and 1.
Typically, the overall similarity is defined as the average of all the individual attribute similarities.
Develop success from failures. Discouragement and failure are two of the surest stepping stones to success.
39

DATA MINING APPLICATIONS
Prediction & Description
Data mining may be used to answer questions like
"would this customer buy a product" or
"is this customer likely to leave?
DM techniques may also be used for sales forecasting and analysis.
Relationship Marketing
Customers have a lifetime value, not just the value of a single sale.
Data mining can help
in analyzing customer profiles and improving direct marketing plans
in identifying critical issues that determine client loyalty and
in improving customer retention
Customer Profiling
This is the process of using the relevant and available information
to describe the characteristics of a group of customers
to identify their discriminators from ordinary consumers and
to identify drivers for their purchasing decisions
This can help an enterprise identify its most valuable customers
so that the enterprise may differentiate their needs and values.
Outliers Identification & Detecting Fraud
For this, examples include:
identifying unusual expense claims by staff
identifying anomalies in expenditure between similar units of an enterprise
identifying fraud involving credit cards
Customer Segmentation
This is a way to assess & view individuals in market based on their status & needs.
Data mining may be used
to understand & predict customer behavior and profitability
to develop new products & services and
to effectively market new offerings
Web site Design & Promotion
Web mining may be used to discover how users navigate a web site and the results
can help in improving the site design.
Web mining may also be used in cross-selling by suggesting to a web customer, items
that he may be interested in.
EXERCISES
1. Explain Data set. Give an Example? (5)
2. Explain 4 types of attributes by giving appropriate example? (10)
3. With example, explain
i) Continious
ii) Discrete
iii) Asymmetric Attributes. (10)
4. Explain general characteristics of data sets. (6)
5. Explain record data & its types. (10)
6. Explain graph based data. (6)
7. Explain ordered data & its types. (10)
8. Explain shortly any five data pre-processing approaches. (10)
9. What is sampling? Explain simple random sampling vs. stratified
sampling vs. progressive sampling. (10)
10. Write a short note on the following: (15)
i)Dimensionality reduction
ii)Variable transformation iii)Feature selection
11. Distinguish between
i) SMC & Jaccard coefficient
ii) Discretization & binarization (6)
12. Explain various distances used for measuring dissimilarity between objects. (6)
13. Consider the following 2 binary vectors
X=(1,0,0,0,0,0,0,0,0,0)
Y=(0,0,0,0,0,0,1,0,0,1)
Find i)hamming distance ii)SMC iii)Jaccard coefficient (4)
14. List out issues in proximity calculation. (4)
15. List any 5 applications of data mining. (8)
Success is the prize for those who stand true to their ideas!
40
UNIT 7: CLUSTERING TECHNIQUES

WHAT IS CLUSTER ANALYSIS?
A cluster is a collection of data-objects
similar to one another within the same group &
dissimilar to the objects in other groups
Cluster analysis is process of grouping a set of data-objects into clusters.
APPLICATIONS OF CLUSTER ANALYSIS

In university, one may wish to find
clusters of students or
clusters of courses
In medicine, one may be interested in finding
clusters of patients or
clusters of diseases
In business, one may want to identify
clusters of customers
clusters of products
Practical applications include
character/pattern recognition
web document classification
image processing
One cannot get through life without pain. What we can do is choose how to use the pain life presents to us.
41

CLASSIFICATION VS. CLUSTER ANALYSIS
Classification is used mostly as a supervised learning method
while clustering is used as unsupervised learning.
Classification
The classes are predefined (Table 4.1).
The user already knows what classes there are.
Some training data that is already labeled by their class-membership is available to build a model.
The classification-problem then is to build a model that would be able to classify newly encountered
data.
Cluster Analysis
One does not know what classes or clusters exist (Table 4.2).
The problem to be solved is to group the given data into meaningful clusters.
The aim of cluster analysis is to find meaningful groups with
small within-group variations &
large between-group variation
Most of the algorithms developed are based on some concept of similarity or distance.
Drawbacks:
This process may be prohibitively expensive for large sets.
Cost of computing distances between groups of objects grows as no. of attributes grows.
Computing distances between categorical attributes is more difficult (compared to
computing distances between objects with numeric attributes)
There are no secrets to success. It is the result of preparation, hard work, and learning from failure.
42

DESIRED FEATURES OF CLUSTER ANALYSIS METHOD
Scalability
Data-mining problems can be large and therefore a cluster-analysis method should be
able to deal with large problems gracefully.
Ideally, performance should be linear with data-size.
The method should also scale well to datasets in which number of attributes is large.
Only one Scan of the Dataset
For large problems, data must be stored on disk, so cost of I/O disk can become
significant in solving the problem.
Therefore, the method should not require more than one scan of disk-resident data.
Ability to Stop & Resume
For large dataset, cluster-analysis may require huge processor-time to complete the
task.
In such cases, task should be able to be stopped & then resumed when convenient.
Minimal Input Parameters
The method should not expect too much guidance from the data-mining analyst.
Therefore, the analyst should not be expected
to have domain knowledge of data and
to posses insight into clusters that might exist in the data
Robustness
Most data obtained from a variety of sources has errors.
Therefore, the method should be able to deal with noise, outlier & missing values
gracefully.
Ability to Discover Different Cluster-Shapes
Clusters appear in different shapes and not all clusters are spherical.
Therefore, method should be able to discover cluster-shapes other than spherical.
Different Data Types
Many problems have a mixture of data types, for e.g. numerical, categorical & even
textual.
Therefore, the method should be able to deal with
numerical data
boolean data
categorical data
Result Independent of Data Input Order
Therefore, the method should not be sensitive to data input-order.
Irrespective of input-order, result of cluster-analysis of the same data should be same.
TYPES OF DATA
Numerical Data
Examples include weight, marks, height, price, salary, and count.
There are a number of methods for computing similarity between these data.
E.g. Euclidean distance, manhattan distance.
Binary Data
Examples include gender, marital status.
A simple method involves
counting how many attribute values of 2 objects are different amongst n
attributes &
using this as an indication of distance
Qualitative Nominal Data
This is similar to binary data which may take more than 2 values but has no natural
order. Examples include religion, foods or colors.
Qualitative Ranked Data
This is similar to qualitative nominal data except that data has an order associated with
it.
Examples include: 1) grades A, B, C, and D
2) sizes S, M, L and XL
One method of computing distance involves transferring the values to numeric values
according to their rank. For example, grades A, B, C, D could be transformed to 4.0, 3.0,
2.0 and 1.0.
Do your work with your whole heart and you will succeed - there's so little competition.
43

COMPUTING DISTANCE
Most cluster-analysis methods are based on measuring similarity between objects.
Distances are normally used to measure the similarity or dissimilarity between 2 objects.
Let the distance between 2 points x and y be D(x,y).
Distance has following simple properties:
1) Distance is always positive. i.e. D(x,y) >= 0
2) Distance from point x to itself is always 0. i.e. D(x,y) = 0
3) Distance from point x to point y is always less than the sum of the distance from x to some
other point z and distance from z to y. i.e. D(x,y) <= D(x,z) + D(z,y)
4) Distance from x to y is always the same as from y to x. i.e. D(x,y) = D(y,x)
Following are some of the distance measures:
1) Euclidean distance (L2 norm of difference vector)
2) Manhattan distance (L1 norm of the difference vector)
3) Chebychev distance (L norm of the difference vector)
4) Categorical data distance
Euclidean Distance
This is most commonly used to compute distances and has an intuitive appeal.
The largest valued attribute may dominate the distance.
Requirement: The attributes should be properly scaled.
This metric is more appropriate when the data is not standardized.
Manhattan Distance
In most cases, the result obtained by this measure is similar to those obtained by using the Euclidean
distance.
The largest valued attribute may dominate the distance.
Chebychev Distance
This metric is based on the maximum attribute difference.
Categorical Data Distance
This metric may be used if many attributes have categorical values with only a small number of
values (e.g. metric binary values).
Let N=total number of categorical attributes.
No matter what we want of life we have to give up something in order to get it.
44

TYPES OF CLUSTER ANALYSIS METHODS
Partitional Method
The objects are divided into non-overlapping clusters (or partitions)
such that each object is in exactly one cluster (Figure 4.1a).
The method obtains a single-level partition of objects.
The analyst has
to specify number of clusters(k) prior
to specify starting seeds of clusters
The analyst may have to use iterative approach in which he has to run the method many times
specifying different numbers of clusters & different starting seeds
then selecting the best solution
The method converges to a local minimum rather than the global minimum.
Figure 4.1a
Figure 4.1b
Hierarchical Methods
A set of nested clusters is organized as a hierarchical tree (Figure 4.1b).
The method either
starts with one cluster & then splits into smaller clusters (called divisive or top down) or
starts with each object in an individual cluster & then tries to merge similar clusters into
larger clusters(called agglomerative or bottom up)
Tentative clusters may be merged or split based on some criteria.
Density based Methods
A cluster is a dense region of points, which is separated by low-density regions, from other regions of
high density.
Typically, for each data-point in a cluster, at least a minimum number of points must exist within a
given radius.
The method can deal with arbitrary shape clusters (especially when noise & outliers are present).
Grid-based methods
The object-space rather than the data is divided into a grid.
This is based on characteristics of the data.
The method can deal with non-numeric data more easily.
The method is not affected by data-ordering.
Model-based Methods
A model is assumed, perhaps based on a probability distribution.
Essentially, the algorithm tries to build clusters with
a high level of similarity within them
a low level of similarity between them.
Similarity measurement is based on the mean values and the algorithm tries to minimize the squared
error function.
Your chances of success in any undertaking can always be measured by your belief in yourself.
45

PARTITIONAL METHODS
The objects are divided into non-overlapping clusters (or partitions)
such that each object is in exactly one cluster (Figure 4.1a).
The method obtains a single-level partition of objects.
The method requires the analyst
to specify number of clusters(k) prior
to specify starting seeds of the clusters
The analyst may have to use iterative approach in which he has to run the method many times
specifying different numbers of clusters & different starting seeds
then selecting the best solution
The method converges to a local minimum rather than the global minimum.
These are popular since they tend to be computationally efficient and are more easily adapted for
very large datasets.
The aim of partitional method is
to reduce the variance within each cluster &
to have large variance between the clusters
THE K-MEANS METHOD
This method can only be used if the data-object is located in the main memory.
The method is called K-means since
each of the K clusters is represented by the mean of the objects( called the centriod) within it.
The method is also called the centroid-method since
at each step, the centroid-point of each cluster is assumed to be known and
each of the remaining points are allocated to the cluster whose centroid is closest to it
The algorithm is as follows
1) Select the number of clusters=k (Figure 4.1c).
2) Pick k seeds as centroids of k clusters. The seeds may be picked randomly unless the user
has some insight into the data.
3) Compute Euclidean distance of each object in the dataset from each of the centroids.
4) Allocate each object to the cluster it is nearest to (based on the distances computed in the
previous step).
5) Compute the centroids of clusters by computing the means of the attribute values of the
objects in each cluster.
6) Check if the stopping criterion has been met (e.g. the cluster-membership is unchanged).
If yes, go to step 7. If not, go to step 3.
7) One may decide
to stop at this stage or
to split a cluster or combine two clusters until a stopping criterion is met
Figure 4.1c
The state of your life is nothing more than a reflection of your state of mind.
46

Example 4.1
Consider the data about students given in Table 4.3.
Step 1 and 2:
Let the three seeds be the first three students as shown in Table 4.4.
Step 3 and 4:
Now compute the distances using the 4 attributes and using the sum of absolute differences for
simplicity. The distance values for all the objects are given in Table 4.5.
Step 5:
Table 4.6 compares the cluster means of clusters found in Table 4.5 with the original seeds.
Do not wait; the time will never be just right. Start where you stand, and work with whatever tools you may have
at your command, and better tools will be found as you go along.
47

Step 3 and 4:
Use the new cluster means to re-compute the distance of each object to each of the means, again
allocating each object to the nearest cluster. Table 4.7 shows the second iteration.
Number of students in cluster 1 is again 2 and the other two clusters still have 4 students each.
A more careful look shows that the clusters have not changed at all. Therefore, the method has
converged rather quickly for this very simple dataset.
The cluster membership is as follows
Cluster C1= {S1, S9}
Cluster C2= {S2, S5, S6, S10}
Cluster C3= {S3, S4, S7, S8}
SCALING AND WEIGHTING
For clustering to be effective, all attributes should be converted to a similar scale.
There are a number of ways to transform the attributes.
One possibility is to transform the attributes
to a normalized score or
to a range(0,1)
Such transformations are called scaling.
Some other approaches are:
1) Divide each attribute by the mean value of that attribute.
This approach reduces the mean of each attribute to 1.
This does not control the variation; some values may still be large, others small.
2) Divide each attribute by the difference between largest-value and smallest-value.
This approach
decreases the mean of attributes that have a large range of values &
increases the mean of attributes that have small range of values.
3) Convert the attribute values to "standardized scores" by subtracting the mean of the
attribute from each attribute value and dividing it by the standard deviation.
Now, the mean & standard-deviation of each attribute will be 0 and 1 respectively.
SUMMARY OF THE K MEANS METHOD
K means is an iterative improvement greedy method.
A number of iterations are normally needed for convergence and
therefore the dataset is processed a number of times.
Following are number of issues related to the method (Disadvantages)
1) The results of the method depend strongly on the initial guesses of the seeds. Need to
specify k, the number of clusters, in advance.
2) The method can be sensitive to outliers. If an outlier is picked as a starting seed, it may endup in a cluster at its own.
3) The method does not consider the size of the clusters.
4) The method does not deal with overlapping clusters.
5) Often, the local optimum is not as good as the global optimum.
6) The method implicitly assumes spherical probability distribution.
7) The method needs to compute Euclidean distances and means of the attribute values of
objects within a cluster. (i.e. Cannot be used with categorical data).
Success is never final and failure never fatal. It is courage that counts.
48

EXPECTATION MAXIMIZATION METHOD
Assumption is that the objects in the dataset have attributes whose values are distributed according
to some linear combination of simple probability distributions.
K-means method involves assigning objects to clusters to minimize within-group variation,
whereas the EM method assigns objects to different clusters with certain probabilities in an
attempt to maximize expectation(or likelihood) of assignment.
The EM method consists of a two-step iterative algorithm:
1) The estimation-step (or E-step) involves estimating the probability distributions of the
clusters given the data.
2) The maximization-step(M-step) involves finding the model parameters that maximize the
likelihood of the solution.
HIERARCHICAL METHODS
A set of nested clusters is organized as a hierarchical tree (Figure 4.1d).
This approach allows clusters to be found at different levels of granularity.
Figure 4.1d
Two types of hierarchical approaches:

1. Agglomerative approach: Each object at the start is a cluster by itself and the nearby
clusters are repeatedly merged resulting in larger clusters until some stopping criterion is met
or all the objects are merged into a single large cluster (Figure 4.1e).
Figure 4.1e
2. Divisive approach: All the objects are put in a single cluster to start. The method then
repeatedly resulting in smaller clusters until a stopping criterion is reached or each cluster has
only one object in it (Figure 4.1f).
Figure 4.1f
It is not the strongest of the species that survives, nor the most intelligent, but the one most responsive to change.
49

DISTANCE BETWEEN CLUSTERS
The hierarchical methods require distances between clusters to be computed. These distance metrics
are often called linkage metrics.
Following methods are used for computing distances between clusters:
1) Single-link(nearest neighbor)
2) Complete-link(farthest neighbor)
3) Centriod
4) Average
5) Ward's minimum variance
SINGLE-LINK
The distance between 2 clusters is defined as the minimum of the distances between all pairs of
points(x,y),
where x is from the first cluster &
y is from the second cluster. D(x, y) = min(xi, yj)
If there are m elements in one cluster and n in another cluster,
all mn pairwise distances must be computed and
the smallest is chosen (Figure 4.2).
Disadvantages:
Each cluster may have an outlier and the 2 outliers may be nearby and so the distance
between the 2 clusters would be computed to be small.
Single link can form a chain of objects as clusters are combined since there is no constraint
on distance between objects that are far away from each other.
COMPLETE-LINK
The distance between 2 clusters is defined as the maximum of the distances between all pairs of
points(x,y). i.e. D(x, y)=max(xi, yj)
This is strongly biased towards compact clusters (Figure 4.3).
Disadvantages:
Each cluster may have an outlier and the 2 outliers may be very far away and so the
distance between the 2 clusters would be computed to be large.
If a cluster is naturally of a shape, say, like a banana then perhaps this method is not
appropriate.
CENTRIOD
The distance between 2 clusters is defined as the distance between the centroids of the clusters.
Usually, the squared Euclidean distance between the centroids is used. i.e. D(x, y)=D(C i, Cj)
Advantages:
This method is easy and generally works well.
This method is more tolerant of somewhat longer clusters than complete link algorithm.
Time is our most valuable asset, yet we tend to waste it, kill it, and spend it rather than invest it.
50

AVERAGE
The distance between 2 clusters is defined as the average of all pairwise distances between
an object from one cluster and
another from the other cluster. i.e. D(x, y)=avg(xi, yj)
Therefore, if there are m elements in one cluster and n in the other,
there are mn distances to be computed, added and divided by mn.
Advantages:
This method tends to join clusters with small variances.
This method is easy and generally works well.
This method is more tolerant of somewhat longer clusters than complete link algorithm.
WARD'S MINIMUM VARIANCE METHOD

Ward's distance is the difference between
total within the cluster sum of squares for the 2 clusters separately and
total within the cluster sum of squares resulting from merging the 2 clusters
An expression for ward's distance is given by
where
DW(A,B) = ward's minimum variance distance between clusters A and B with NA & NB objects in
them respectively.
Dc(A,B) = centriod distance between the 2 clusters computed as squared Euclidean distance
between the centroids.
Advantages:
This method generally works well and results in creating small tight clusters
This method produces clusters with roughly the same number of objects
This method tends to join clusters with a small number of objects
The distance measure can be sensitive to outliers
Pay no attention to what the critics say; there has never been set up a statue in honor of a critic.
51

AGGLOMERATIVE METHOD
This method is basically a bottom-up approach.
The algorithm is as follows:
1) Allocate each point to a cluster of its own. Thus, we start with n clusters for n objects.
2) Create a distance-matrix by computing distances between all pairs of clusters (either using
the single link metric or the complete link metric). Sort these distances in ascending order.
3) Find the 2 clusters that have the smallest distance between them.
4) Remove the pair of objects and merge them.
5) If there is only one cluster left then stop.
6) Compute all distances from the new cluster and update the distance-matrix after the merger
and go to step 3.
Example 4.3
We now use agglomerative technique for clustering the data given in Table 4.10.
Steps 1 and 2:
Allocate each point to a cluster and compute the distance-matrix using the centroid method.
The distance-matrix is symmetric, so we only show half of it in Table 4.11.
Table 4.11 gives the distance of each object with every other object.
Steps 3 and 4:
The smallest distance is 8 between objects S 4 and S8. They are combined and removed and we put
the combined cluster (C1) where the object S4 was.
Table 4.12 is now the new distance-matrix. All distances except those with cluster C 1 remain
unchanged.
Our very business in life is not to get ahead of others... but to get ahead of ourselves.
52
Steps 5 and 6:
The smallest distance now is 15 between objects S 5 and S6. They are combined in C2 cluster and S5
and S6 are removed.
Steps 3, 4, 5 and 6: Table 4.13 is the updated distance-matrix.
The result of using the agglomerative method could be something like that shown in Figure 4.6.
Happy people plan actions, they don't plan results.

53

DIVISIVE HIERARCHICAL METHODS
These methods
start with the whole dataset as one cluster
then proceed to recursively divide the cluster into two sub-clusters and
continue until each cluster has only one object.
Two types of divisive methods are:
1) Monothetic: This splits a cluster using only one attribute at a time.
An attribute that has the most variation could be selected.
2) Polythetic: This splits a cluster using all of the attributes together.
Two clusters far apart could be build based on distance between objects.
1) Decide on a method of measuring the distance between 2 objects. Also, decide a threshold
distance.
2) Create a distance-matrix by computing distances between all pairs of objects within the
cluster. Sort these distances in ascending order.
3) Find the 2 objects that have the largest distance between them. They are the most dissimilar
objects.
4) If the distance between the 2 objects is smaller than the pre-specified threshold and there is
no other cluster that needs to be divided then stop, otherwise continue.
5) Use the pair of objects as seeds of a K-means method to create 2 new clusters.
6) If there is only one object in each cluster then stop otherwise continue with step 2.
We need to resolve the following 2 issues:
1) Which cluster to split next?
i) Split the cluster in some sequential order.
ii) Split the cluster that has the largest number of objects.
iii) Split the cluster that has the largest variation within it.
2) How to split a cluster?
A distance-matrix is created and the 2 most dissimilar objects are selected as seeds
of 2 new clusters. The K-means method is then used to split the cluster.
Example 4.4
Consider the distance-matrix in Table 4.14.
The largest distance is 115 between objects S 8 and S9. They become the seeds of 2 new clusters.
K means is used to split the group into 2 clusters.
Cluster C1 includes S4, S7, S8 and S10.

Cluster C2 includes S1, S2, S3, S5, S6, S9.
Since none of the stopping criteria have been met, we decide to split the larger cluster next and then
repeat the process.
We find the largest distance in C2 first as shown in Table 4.16. The largest distance in C 2 is 82
between S3 & S6. C2 can therefore be split with S3 & S6 as seeds.
When your desires are strong enough you will appear to possess superhuman powers to achieve.
54
The distance-matrix of C1 is given in Table 4.17. The largest distance is 98 between S 8 and S10. C1
can therefore be split with S8 and S10 as seeds.
The method continues like this until the stopping criteria is met.
SUMMARY OF HIERARCHICAL METHODS
Advantages
1) This method is conceptually simpler and can be implemented easily.
2) In some applications, only proximity-data is available and then this method may be better.
3) This method can provide clusters at different levels of granularity.
4) This method can provide more insight into data by showing a hierarchy of clusters (than a
flat cluster structure created by a partitioning method like the K-means method).
5) Do not have to assume any particular number of clusters.
Disadvantage
1) Do not scale well: Time complexity of at least O(n2), where n is number of total objects.
2) The distance-matrix requires O(n2) space and becomes very large for a large number of
objects.
3) Different distance metrics and scaling of data can significantly change the results.
4) Once a decision is made to combine two clusters, it cannot be undone.
If you dont set goals for yourself, you are doomed to work to achieve the goals of someone else.
55

DENSITY-BASED METHODS
A cluster is a dense region of points, which is separated by low-density regions, from other regions of
high density.
Typically, for each data-point in a cluster, at least a minimum number of points must exist within a
given radius.
Data that is not within such high-density clusters is regarded as outliers or noise.
DBSCAN is one example of a density-based method for clustering.
DBSCAN stands for Density Based Spatial Clustering of Applications with Noise.
This requires 2 input parameters:
size of the neighborhood(R) &
minimum points in the neighborhood(N).
The point-parameter N
determines the density of acceptable clusters &
determines which objects will be labeled outliers or noise.
The size-parameter R determines the size of the clusters found.
If R is big enough, there would be one big cluster and no outliers.
If R is small, there will be small dense clusters and there might be many outliers.
We define a number of concepts that are required in the DBSCAN method:
1) Neighbourhood: The neighborhood of an object y is defined as all the objects that are within
the radius R from y.
2) Core-object: An object y is called a core-object if there are N objects within its neighborhood.
3) Proximity: Two objects are defined to be in proximity to each other if they belong to the
same cluster. Object x1 is in proximity to object x2 if two conditions are satisfied:
i) The objects are close enough to each other, i.e. within a distance of R.
ii) x2 is a core object.
4) Connectivity: Two objects x1 and xn are connected if there is a chain of objects x1,x2. . . .xn
from x1 to xn such that each xi+1 is in proximity to object xi.
The algorithm is as follows
1) Select values of R and N (Figure 4.7).
2) Arbitrarily select an object p.
3) Retrieve all objects that are connected to p, given R and N.
4) If p is a core object, a cluster is formed.
5) If p is a border object, no objects are in its proximity. Choose another object. Go to step 3.
6) Continue the process until all of the objects have been processed.
Figure 4.7: DBSCAN
Plan for the future, because that is where you are going to spend the rest of your life.
56

DEALING WITH LARGE DATABASES
A method requiring multiple scans of data that is disk-resident could be quite inefficient for large
problems.
K MEANS METHOD FOR LARGE DATABASES
The method
first picks the number of clusters & their seed centriods and
then attempts to classify each object to belong to one of the following 3 groups:
1) Those that are certain to belong to a cluster. These objects together are called discard-set.
Some information about these objects is computed and saved. This includes
number of objects n
a vector sum of all attribute values of the n objects and
a vector sum of squares of all attribute values of n objects
2) Those that are sufficiently close to each other to be replaced by their summary. These
objects together are called the compression-set.
The objects are however sufficiently far away from each cluster's centroid that they
cannot be put in the discard set.
3) The remaining objects that are too difficult to assign to either of the 2 groups above. These
objects are called the retained-set.
These are stored as individual objects.
They cannot be replaced by a summary.
HIERARCHICAL METHOD FOR LARGE DATABASES - CONCEPT OF FRACTIONATION
Main idea is as follows:
1) First split the data into manageable subsets called fractions &
2) Then apply a hierarchical method to each fraction. The concept is called fractionation.
Let M = largest number of objects that the hierarchical method may be applied to.
The size M may be determined based on the size of the main memory.
1) Split the large dataset into fractions of size M.
2) The hierarchical method is applied to each fraction.
Let C = number of clusters obtained from all the fractions.
3) For each cluster, compute the mean of the attribute values of the objects.
Let this mean vector be mi, i=1, 2, . . C.
These cluster means are called meta-observation.
The meta-observation now becomes the data values that represent the fractions.
4) If the C meta-observations are too large, go to step 1,
otherwise apply same hierarchical method to the meta-observations obtained in step 3.
5) Allocate each object of the original dataset to the cluster with the nearest mean obtained in
step 4.
CLUSTER ANALYSIS SOFTWARE
ClustanGraphics7 from Clustan offers a variety of clustering methods including K means, density
based and hierarchical cluster analysis. The software provides facilities to display results of clustering
including dendograms and scatterplots.
CViz Cluster visualization from IBM is a visualization tool designed for analyzing high dimensional
data in large, complex data set.
Cluster 3.0, open source software. This uses the k means method, which includes multiple trials to
find the best clustering solution.
CLUTO provides a set of clustering methods including partitional, agglomerative and graph
partitioning based on a variety of similarity/distance metrics.
Nothing is particularly hard if you divide it into small jobs.

57

QUALITY AND VALIDITY OF CLUSTER ANALYSIS METHODS
Let number of clusters = k
Let clusters = Ci, i=1... . . . . k
Let total number of objects = N
Let number of objects in cluster = Mi
so that
The within-cluster variation between the objects in a cluster is defined as the average squared
distance of each object from the centroid of the cluster.
If mi is the centroid of the cluster Ci
then the mean of the cluster is given by
and the internal cluster variation is given by
The average within-cluster variation is given by
The between cluster distances E is the average sum of squares of pairwise distances between the
centroids of the k clusters. We may write E as
To achieve the best result of cluster analysis,

one possible approach might be choosing result that has largest E/I from the results available.
If E is large, it shows good separation between the clusters.
If I is small, it means that we have tight clusters.
The quality of a clustering method involves a number of criteria
1) Efficiency of the method.
2) Ability of the method to deal with noisy and missing data.
3) Ability of the method to deal with large problems.
4) Ability of the method to deal with a variety of attributes types and magnitudes.
EXERCISES
1) What is cluster analysis? What are its applications? (2)
2) Compare classification vs. cluster analysis. (6)
3) List out and explain desired features of cluster analysis method. (6)
4) List out and explain different types of data. (4)
5) List out and explain different distance measures. (4)
6) List out and explain different types of cluster analysis methods. (6)
7) Write algorithm for k-means method. (6)
8) Apply k-means method for clustering the data given in Table 4.3. (6)
9) List out disadvantages of k-means method. (6)
10) Explain scaling and weighting. (4)
11) Explain expectation maximization method. (4)
12) Compare agglomerative approach vs. divisive approach. (4)
13) Explain different methods used for computing distances b/t clusters. (6)
14) Write algorithm for agglomerative approach. (6)
15) Apply agglomerative technique for clustering data given in Table4.10. (6)
16) Write algorithm for divisive approach. (6)
17) List out advantages and disadvantages of hierarchical methods. (6)
18) Explain DBSCAN with its algorithm. (6)
19) Explain K means method for large databases. (4)
20) Explain hierarchical method for large databases. (6)
21) Explain quality and validity of cluster analysis methods (6)
A man is a success if he gets up in the morning and goes to bed at night and in between does what he wants to do.
58
UNIT 8: WEB MINING

WEB MINING
Web-mining is the application of data-mining techniques to extract knowledge from web-data. (i.e.
web-content, web-structure, and web-usage data).
We interact with the web for the following purposes:
1) Finding Relevant Information
We use the search-engine to find specific information on the web.
Query triggered process: We specify a simple keyword-query and the response from a searchengine is a list of pages, ranked by their similarity to the query.
Search tools have the following problems:
i) Low precision: This is due to the irrelevance of many of the search-results.
We may get many pages of information which are not really relevant to our query.
ii) Low recall: This is due to inability to index all the information available on the web.
Because some of the relevant pages are not properly indexed.
2) Discovering New Knowledge from the Web
Data triggered process: This assumes that
we already have a collection of web-data and
we want to extract potentially useful knowledge out of it
3) Personalized Web Page Synthesis
We may wish to synthesize a web-page for different individuals from the available set of webpages.
While interacting with the web, individuals have their own preferences for the style of the
content and presentation.
4) Learning about Individual Users
This is about knowing
what the customers do and
what the customers want
Within this problem, there are sub-problems such as
problems related to effective web-site design and management
problems related to marketing etc
Techniques from web-mining can be used to solve these problems.
Other related techniques from different research areas, such as DB(database),
IR(information retrieval) & NLP(natural language processing), can also be used.
Web-mining has 3 main operations:
clustering (e.g. finding natural groupings of users, pages)
associations (e.g. which URLs tend to be requested together)
sequential analysis (e.g. the order in which URLs tend to be accessed)
Web mining techniques can be classified into 3 areas of interest (Figure 10.1):
web-content mining (e.g. text, image, records, etc.)
web-structure mining (e.g. hyperlinks, tags, etc)
web-usage mining (e.g. http logs, app server logs, etc.)
Figure 10.1: Web mining
Without discipline, there's no life at all.

59

WEB CONTENT MINING
This is the process of extracting useful information from the contents of web-documents.
We see more & more government information are gradually being placed on the web in recent years.
We have
digital libraries which users can access from the web
web-applications which users can access through web-interfaces
Some of the web-data are hidden-data, and
some are generated dynamically as a result of queries and reside in the DBMSs.
The web-content consists of different types of data such as text, image, audio, video as well as
hyperlinks.
Most of the research on web-mining is focused on the text or hypertext contents.
The textual-parts of web-data consist of
unstructured-data such as free texts
semi structured-data such as HTML documents &
structured-data such as data in the tables
Much of the web-data is unstructured, free text-data.
As a result, text-mining techniques can be directly employed for web-mining.
Issues addressed in text mining are, topic discovery, extracting association patterns, clustering of
web documents and classification of Web Pages.
Research activities on this topic have drawn heavily on techniques developed in other disciplines such
as IR(Information Retrieval) and NLP(Natural Language Processing).
WEB USAGE MINING
This deals with studying the data generated by the web-surfer's sessions (or behaviors).
Web-content/structure mining utilise the primary-data(or real) on the web.
On the other hand, web-usage mining extracts the secondary-data derived from the
interactions of the users with the web.
The secondary-data includes the data from
web-server access logs
browser logs
user transactions/queries
user profiles
registration/bookmark data
cookies
There are 2 main approaches in web-usage mining:
1) General Access Pattern Tracking
This can be used to learn user-navigation patterns.
This can be used to analyze the web-logs to understand access-patterns and trends.
This can shed better light on the structure & grouping of resource providers.
2) Customized Usage Tracking
This can be used to learn a user-profile in adaptive interfaces (personalized).
This can be used to analyze individual trends.
Main purpose: is to customize web-sites to users.
Based on user access-patterns, following things can be dynamically customized for
each user over time:
information displayed
depth of site-structure
format of resources
The mining techniques can be classified into 2 commonly used approaches:
1) The first approach maps the usage-data of the web-server into relational-tables before a
traditional data-mining technique is applied.
2) The second approach uses the log-data directly by utilizing special pre-processing
techniques.
You can do what you think you can do and you cannot do what you think you cannot do.
60

WEB STRUCTURE MINING
The structure of a typical web-graph consists of web-pages as nodes, and hyperlinks as edges
connecting related pages.
Web-Structure mining is the process of discovering structure information from the web.
This type of mining can be performed either at the (intra-page) document level or at the (interpage) hyperlink level.
This can be used to classify web-pages.
This can be used to generate information such as the similarity & relationship between different
web-sites.
PageRank
PageRank is a metric for ranking hypertext documents based on their quality.
The key idea is that a page has a high rank if it is pointed to by many highly ranked pages
The PageRank of a page A is given by
Here, n = number of nodes in the graph and
OutDegree(q) = number of hyperlinks on page q
d = damping factor which can be set between 0 and 1 and is usually set to 0.85
Clustering & Determining Similar Pages
For determining the collection of similar pages, we need to define the similarity measure
between the pages. There are 2 basic similarity functions:
1) Co-citation
For a pair of nodes p and q, the co-citation is the number of nodes that point to both p
and q.
2) Bibliographic coupling
For a pair of nodes p and q, the bibliographic coupling is equal to the number of nodes
that have links from both p and q.
Social Network Analysis
This can be used to measure the relative standing or importance of individuals in a network.
The basis idea is that if a web-page points a link to another web-page, then the former is,
in some sense, endorsing the importance of the latter.
Links in the network may have different weights, corresponding to the strength of
endorsement.
People with small minds talk about other people. People with average minds talk about events.
People with great minds talk about ideas,
61

TEXT MINING
This is concerned with various tasks, such as
extraction of information implicitly contained in the collection of documents or
similarity-based structuring
Text-collection lacks the imposed structure of a traditional database.
The text expresses a vast range of information, but encodes the information in a form that is
difficult to decipher automatically.
Traditional data-mining techniques have been designed to operate on structured-databases.
In structured-databases, it is easy to define the set of items and hence, it becomes easy to employ
the traditional mining techniques.
In textual-database, identifying individual items (or terms) is not so easy.
Text-mining techniques have to be developed to process the unstructured textual-data to aid in
knowledge discovery.
The inherent nature of textual-data, namely unstructured characteristics, motivates the
development of separate text-mining techniques.
Following approaches can be used for text-mining:
1) One way is to impose a structure on the textual-database and use any of the known datamining techniques meant for structured-databases.
2) The other approach would be to develop a very specific technique for mining that exploits
the inherent characteristics of textual-databases.
INFORMATION RETRIEVAL (IR)
This is concerned with finding and ranking documents that match the users' information needs.
The way of dealing with textual-information is a keyword-based document representation.
A body of text is analyzed by its constituent-words, and various techniques are used to build the
core words for a document.
The goals are
to find documents that are similar based on some specification of the user
to find right index terms in a collection, so that querying will return appropriate document
Recent trends in IR research include
document classification
data visualization
filtering
INFORMATION EXTRACTION (IE)
This is concerned with transforming a collection of documents, usually with the help of an IR
system, into information that is more readily digested and analyzed.
IE extracts relevant facts from the documents
whereas IR selects relevant documents.
In general, IE works at a finer granularity level than IR does on the documents.
Most IE systems use data-mining( or machine-learning) techniques to learn the extraction patterns
(or rules) for documents semi-automatically or automatically.
The results of the IE process could be in the form of
structured database or
summary/compression of the original document
One important key to success is self-confidence. An important key to self- confidence is preparation.
62

UNSTRUCTURED TEXT
Unstructured-documents are free texts, such as news stories.
To convert an unstructured-document to a structured form, following features can be extracted:
Word Occurrences
The vector-representation (or bag of words) takes single words found in the
training-set as features, ignoring the sequence in which the words occur.
The feature is said to be boolean, if we consider whether a word either occurs or
does not occur in a document.
The feature is said to be frequency-based, if the frequency of the word in a
document is taken into consideration.
Stopword Removal
Stopwords are frequently occurring and insignificant words in a language that help
construct sentences but do not represent any content of the documents.
Common stopwords in English include:
a, about, an, are, as, at, be, by, for, from, how, in, is, of, on, or,
that, the, these, this, to, was, what, when, where, who, will, with
Such words can be removed before documents are indexed and stored.
Stemming
Stemming refers to the process of reducing words to their morphological roots or
stems
A stem is the portion of a word that is left after removing its prefixes and suffixes.
For example, "informing", "information", "informer" & "informed" are reduced to
"inform".
POS
POS stands for Part of Speech.
There can be 25 possible values for POS tags.
Most common tags are noun, verb, adjective and adverb.
Thus, we can assign a number 1,2,3,4 or 5, depending on whether the word is a
noun, verb, adjective, adverb or any other, respectively.
Positional Collocations
The values of this feature are the words that occur one or two position to the right or
left of the given word.
n-gram
An n-gram is a contiguous sequence of n items from a given sequence of text.
This can be used for predicting the next item in a sequence.
LSI
LSI stands for Latent Semantic Indexing.
LSI is an indexing and retrieval method to identify the relationships between the
terms and concepts contained in an unstructured collection of text.
The one without dreams is the one without wings.

63

TEXT CLUSTERING
Once the features of an unstructured-text are identified or the structured-data of the text is
available, text-clustering can be done by using any clustering technique.
One popular text-clustering algorithm is ward's minimum variance method.
Wards method is an agglomerative hierarchical clustering technique and it tends to generate very
compact clusters.
We can take either the Euclidean metric or hamming distance as the measure of the dissimilarities
between feature vectors.
The clustering method begins with n clusters, one for each text.
At any stage, 2 clusters are merged to generate a new cluster based on the following criterion:
where xk is mean value of the dissimilarity for the cluster Ck and

nk is the number of elements in this cluster.
Scatter/Gather
This is a method of grouping the documents based on the overall similarities in their content.
Scatter/gather is so named because it allows the user to scatter documents into groups(or
clusters), then gather a subset of these groups and re-scatter them to form new groups.
Each cluster is represented by a list of topical terms, that is, a list of words that attempt to give
the user an idea of what the documents in the cluster are about.
EXERCISES
Write a short note on following (5 marks each)
1) Web mining
2) Web content mining
3) Web structure mining
4) Web usage mining
5) Text mining
6) IR and IE
7) Unstructured text
8) Text clustering
The future belongs to those who believe in the beauty of the dream.
64
UNIT 8(CONT.): TEMPORAL & SPATIAL DATA MINING

TEMPORAL DATA MINING
This can be defined as non-trivial extraction of potentially-useful & previously-unrecorded
information with an implicit/explicit temporal-content, from large quantities of data.
This has the capability to infer causal and temporal-proximity relationships.
Consider, for example, an association-rule - "Any person who buys a car also buys a steeringlock". But if we take the temporal-aspect into consideration, this rule would be - "Any person who
buys a car also buys a steering-lock after that".
TYPES OF TEMPORAL DATA
There can be 4 different levels of temporality:
Static
Static-data are free of any temporal-reference.
Inferences derived from static-data are also free of any temporality.
Sequences (Ordered Sequences of Events)
Though there may not be any explicit reference to time,
there exists a temporal-relationship between data-items.
For example, market-basket transaction. The entry-sequence of transactions
automatically incorporates a sort of temporality.
While most collections are often limited to the sequence-relationships before and
after, this category also includes the richer relationships , such as during, meet,
overlap etc.
Sequence-mining is one of the major activities in temporal-mining.
Timestamped
The temporal-information is explicit.
The relationship can be quantitative, in the sense that
we can say the exact temporal-distance between the data-elements &
we can say that one transaction occurred before another
For example, census data, land-use data etc.
Inferences derived from this data can be temporal or non-temporal.
Time-series data are a special case of this category. The events are uniformly
spaced on the time-scale.
Fully Temporal
The validity of the data-element is time-dependent.
Inferences derived from this data are necessarily temporal.
TEMPORAL DATA MINING TASKS
There can be 4 different tasks:
Temporal Association
Here, we attempt to discover temporal-associations between non-temporal itemsets.
For example, "70% of the readers who buy a DBMS book also buy a Data Mining
book after a semester".
Temporal Classification
We can extend the concept of decision-tree construction on temporal-attributes.
For example, a rule could be: "The first case of malaria is normally reported after
the first pre-monsoon rain and during the months of May-August".
Trend Analysis
The analysis of one or more time series of continuous data may show similar trends
i.e. similar shapes across the time axis.
For example, "The deployment of the Android OS is increasingly becoming popular in
the smartphone industry".
If you want to reach a goal, you must "see the reaching" in your own mind before you actually arrive at your goal.
65

TEMPORAL ASSOCIATION RULES
Association rules identify whether a particular subset of items is supported by an adequate number
of transactions.
The presence of a temporal association rule may suggest a number of interpretations such as
The earlier event plays some role in causing the later event
There is a third set of reasons that cause both events
The confluence of events is coincidental
Temporal association rules are sometimes viewed in the literature as causal rules.
Causal rules describe relationships, where changes in one event cause subsequent changes in
other parts of the domain.
They are common targets of scientific investigation within the medical domain, where the search
for factors that may cause or aggravate a particular disease is important.
The static properties, such as gender, and the temporal properties, such as medical treatments,
are taken into account during mining.
SPATIAL DATABASE
A spatial database stores a large amount of space-related data(or objects), such as
map navigation data
preprocessed remote sensing data or
medical imaging data
Spatial objects are objects that occupy space.
Spatial objects usually consist of both spatial and non-spatial data.

Spatial data: data related to spatial description of the objects such as coordinates, areas,
latitudes, perimeters, spatial relations (distance, topology, direction).
Example: earthquake points, town coordinates on map, etc.
Non-spatial data: other data associated to spatial objects.
Example: earthquake degrees, population of a town, etc
Spatial objects can be classified into different classes (hospital class, airport class, etc.), several
classes can be hierarchically organized in layers (countries, provinces, districts, etc.)
Non-spatial data can be stored in relational databases
Spatial data are typically described by
Geometric properties: coordinates(schools), lines(roads, rivers), area(countries, cites)
Relational properties: adjacency(A is neighbor of B), inclusion(A is inside in B)
Applications and Problems:
GIS(Geographic information systems) store information related to geographic locations on
Earth (Weather, community infrastructure needs, disaster management)
Homeland security issues such as prediction of unexpected events & planning of evacuation.
Remote sensing and image classification
Biomedical applications include medical imaging and illness diagnosis
It is literally true that you can succeed best and quickest by helping others to succeed.
66

SPATIAL DATA MINING
This refers to the extraction of knowledge, spatial relationships, or other interesting patterns not
explicitly stored in spatial-databases.
Consider a map of the city of Mysore containing various natural and man-made geographic
features, and clusters of points (where each point marks the location of a particular house).
The houses might be important because of their size, or their current market value.
Clustering algorithms can be used to assign each point to exactly one cluster, with the number of
clusters being defined by the user.
We can mine varieties of information by identifying likely relationships.
For ex, "the land-value of cluster of residential area around Mysore Palace is high".
Such information could be of value to realtors, investors, or prospective home buyers.
This problem is not so simple because there may be a large number of features to consider.
We need to be able to detect relationships among large numbers of geo-referenced objects without
incurring significant overheads.
SPATIAL MINING TASKS
This includes
finding characteristics rules
discriminant rules
association rules
A spatial-characteristic rule is a general description of spatial-data. For example, a rule describing
the general price ranges of houses in various geographic regions in a city.
A spatial-discriminant rule is a general description of the features discriminating a class of spatialdata from other classes, For example, the comparison of price range of houses in different
geographical regions.
Spatial Association Rles
These describe the association between spatially related objects.
We can associate spatial attributes with non spatial attributes. For example, "the monthly
rental of houses around the market area is mostly Rs 500 per sq mt."
We can associate spatial attributes with spatial attributes. For example, "uncontrolled
tapping of borewell water in the market area may most likely cause earthquake in the
neighbouring areas".
Attribute-oriented Induction
The concept hierarchies of spatial and non-spatial attributes can be used to determine
relationships between different attributes.
One may be interested in a particular category of land-use patterns.
A built-up area may be a recreational facility or a residential complex.
Similarly, a recreational facility may be a cinema or a restaurant.
One has to be very specific regarding the level of details to which one wants to discover
spatial knowledge.
Aggregate Proximity Relationships
This problem is concerned with relationships between spatial-clusters based on spatial and
non-spatial attributes.
Given n input clusters, we want to associate the clusters with classes of features (e.g.
educational institutions which, in turn, may be comprised of secondary schools and junior
colleges or higher institutions).
The problem is to find classes of features that are in close proximity to most(or all) of the
spatial clusters.
Your attitude, not your aptitude, will determine your altitude.

67

SPATIAL CLUSTERING
The key idea of a density based cluster is that for each point of a cluster, its espilon neighbourhood
has to contain at least a minimum number of points.
We can generalize this concept in 2 different ways:
1) First, any other symmetric & reflexive neighbourhood relationship can be used instead of
an espilon neighbourhood. It may be more appropriate to use topological relations such as
intersects, meets or above/below to group spatially extended objects.
2) Second, instead of simply counting the objects in a neighbourhood of an object, other
measures to define the "cardinality" of that neighbourhood can be used as well.
Spatial Characterization
A spatial-characterization is a description of the spatial and non-spatial properties, which
are typical for the target-objects but not for the whole database.
For instance, different object types in a geographic database are mountains, lakes,
highways, railroads etc.
Spatial characterization considers both
properties of the target objects &
properties of their neighbours
A spatial characterization rule of the form - "Apartments in Sainikpur have a high
occupancy rate of retired army officers"- is an example.
SPATIAL TREND
This can be defined as a regular change of one or more non-spatial attributes, when moving away
from a given spatially referenced-object.
For instance, "when we move away northwards from the 'Highway Circle, the rentals of residential
houses decreases approximately at the rate of 5% per kilometer".
One can also show the spatial trends pictorially by overlaying the direction of trend on a map.
SPATIO TEMPORAL DATA MINING
A spatiotemporal database is a database that manages both space and time information.
Common examples include:
Tracking of moving objects, which typically can occupy only a single position at a given time.
A database of wireless communication networks, which may exist only for a short timespan
within a geographic region.
Spatio-temporal data mining is an emerging research area dedicated to the development and
application of novel computational techniques for the analysis of large spatiotemporal databases
This encompasses techniques for discovering useful spatial and temporal relationships or patterns
that are not explicitly stored in spatiotemporal datasets.
Both the temporal and spatial dimensions add substantial complexity to data mining tasks.
Classical data mining techniques often perform poorly when applied to spatiotemporal data sets for
many reasons.
First, spatial data is embedded in a continuous space, whereas classical datasets are in discrete
notions like transactions.
Second, a common assumption about independence of data samples in classical statistical analysis
is generally false because spatial data tends to be highly auto-correlated. (For example, people with
similar characteristics, occupation and background tend to cluster together in the same
neighborhoods).
EXERCISES
Write a short note on following (5 marks each)
1) Temporal data mining
2) Temporal data mining tasks
3) Temporal association rules
4) Spatial mining
5) Spatial mining tasks
6) Spatial clustering
7) Spatial trend
8) Spatio Temporal data mining
The deepest human need is the need to be appreciated.

68

Vtu 7TH Sem Cse/ise Data Warehousing & Data Mining Notes 10cs755/10is74

Uploaded by

Copyright:

Vtu 7TH Sem Cse/ise Data Warehousing & Data Mining Notes 10cs755/10is74

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Vtu 7TH Sem Cse/ise Data Warehousing & Data Mining Notes 10cs755/10is74

Uploaded by

Copyright:

What are the main topics covered in the document?

What are the main topics covered in the document?

What are some examples of spatial characterization rules mentioned?

What are some examples of spatial characterization rules mentioned?

DATA WAREHOUSING AND DATA MINING

Subject Code: 10IS74/10CS755

UNIT 2: ONLINE ANALYTICAL PROCESSING (OLAP)

UNIT 3: DATA MINING

UNIT 7: CLUSTERING TECHNIQUES

UNIT 8: WEB MINING/TEMPORAL & SPATIAL DATA MINING

DATA WAREHOUSING & DATA MINING

UNIT 1: DATA WAREHOUSING

DATA WAREHOUSING & DATA MINING

DATA WAREHOUSING & DATA MINING

DATA WAREHOUSING & DATA MINING

Happiness is our birthright.

DATA WAREHOUSING & DATA MINING

DATA WAREHOUSING & DATA MINING

ODS & DW ARCHITECTURE

DATA WAREHOUSING & DATA MINING

If you can concentrate only on the present, youll be a happy man.

DATA WAREHOUSING & DATA MINING

When you know better, you can do better.

DATA WAREHOUSING & DATA MINING

DATA WAREHOUSING & DATA MINING

DATA WAREHOUSING & DATA MINING

UNIT 2: ONLINE ANALYTICAL PROCESSING (OLAP)

MOTIVATIONS FOR USING OLAP

DATA WAREHOUSING & DATA MINING

DATA WAREHOUSING & DATA MINING

DATA WAREHOUSING & DATA MINING

Learning should be fun, memorable and moving.

DATA WAREHOUSING & DATA MINING

Experience is always the best teacher.

DATA WAREHOUSING & DATA MINING

Figure 8.2: cube formed by tables 8.6, 8.7 & 8.8

DATA WAREHOUSING & DATA MINING

DATA WAREHOUSING & DATA MINING

DATA WAREHOUSING & DATA MINING

DATA WAREHOUSING & DATA MINING

DATA WAREHOUSING & DATA MINING

Life shrinks or expands in proportion to one's courage.

DATA WAREHOUSING & DATA MINING

DATA WAREHOUSING & DATA MINING

UNIT 3: DATA MINING

Difficulties increase the nearer we approach the goal.

DATA WAREHOUSING & DATA MINING

DATA WAREHOUSING & DATA MINING

DATA WAREHOUSING & DATA MINING

DATA WAREHOUSING & DATA MINING

UNIT 3: DATA MINING (CONT.)

PROPERTIES OF ATTRIBUTE VALUES

When the promise is clear, the price gets easy.

DATA WAREHOUSING & DATA MINING

DESCRIBING ATTRIBUTES BY THE NUMBER OF VALUES

DATA WAREHOUSING & DATA MINING

DATA WAREHOUSING & DATA MINING

TYPES OF RECORD DATA

DATA WAREHOUSING & DATA MINING

No one knows what he can do until he tries.

DATA WAREHOUSING & DATA MINING

Do not let the future be held hostage by the past.