Data Mining: Concepts and Techniques.ppt

Data Mining:
Concepts and Techniques
Unit 1

Outline
 1.1 Motivation: Why data mining?
 1.2 What is data mining?
 1.3 Data Mining: On what kind of data?
 1.4 Data mining functionality: What kinds of Patterns Can Be Mined?
 1.5 Are all the patterns interesting?
 1.6 Classification of data mining systems
 1.7 Data Mining Task Primitives
 1.8 Integration of data mining system with a DB and DW System
 1.9 Major issues in data mining
Data Mining: Concepts and Techniques

1.1 Why Data Mining?
 The Explosive Growth of Data: from terabytes(10004
) to yottabytes(10008
)
 Data collection and data availability
 Automated data collection tools, database systems, web
 Major sources of abundant data
 Business: Web, e-commerce, transactions, stocks, …
 Science: bioinformatics, scientific simulation, medical research …
 Society and everyone: news, digital cameras, …
 Data rich but information poor!
 What does those data mean?
 How to analyze data?
 Data mining — Automated analysis of massive data sets

Evolution of Database Technology

1.2 What Is Data Mining?
 Data mining (knowledge discovery from data)
 Extraction of interesting (non-trivial, implicit, previously
unknown and potentially useful) patterns or knowledge from
huge amount of data
 Data mining: a misnomer?
 Alternative names
 Knowledge discovery (mining) in databases (KDD),
knowledge
extraction, data/pattern analysis, data archeology, data
dredging, information harvesting, business intelligence, etc.
5

Potential Applications
 Data analysis and decision support
 Market analysis and management
 Target marketing, customer relationship management (CRM),
market basket analysis, cross selling, market segmentation
 Risk analysis and management
 Forecasting, customer retention, improved underwriting, quality
control, competitive analysis
 Fraud detection and detection of unusual patterns (outliers)
 Other Applications
 Text mining (news group, email, documents) and Web mining
 Stream data mining
 Bioinformatics and bio-data analysis

Ex.: Market Analysis and Management
 Where does the data come from?—Credit card transactions, loyalty cards,
discount coupons, customer complaint calls, surveys …
 Target marketing
 Find clusters of “model” customers who share the same characteristics: interest,
income level, spending habits, etc.,
 E.g. Most customers with income level 60k – 80k with food expenses $600 - $800 a month live in that area
 Determine customer purchasing patterns over time
 E.g. Customers who are between 20 and 29 years old, with income of 20k – 29k usually buy this type of CD
player
 Cross-market analysis—Find associations/co-relations between product
sales, & predict based on such association
 E.g. Customers who buy computer A usually buy software B
7

Ex.: Market Analysis and Management (2)
 Customer requirement analysis
 Identify the best products for different customers
 Predict what factors will attract new customers
 Provision of summary information
 Multidimensional summary reports
 E.g. Summarize all transactions of the first quarter from three different branches
Summarize all transactions of last year from a particular branch
Summarize all transactions of a particular product
 Statistical summary information
 E.g. What is the average age for customers who buy product A?
 Fraud detection
 Find outliers of unusual transactions
 Financial planning
 Summarize and compare the resources and spending
8

Knowledge Discovery (KDD) Process

KDD Process: Several Key Steps
 Learning the application domain
 relevant prior knowledge and goals of application
 Identifying a target data set: data selection
 Data processing
 Data cleaning (remove noise and inconsistent data)
 Data integration (multiple data sources maybe combined)
 Data selection (data relevant to the analysis task are retrieved from database)
 Data transformation (data transformed or consolidated into forms appropriate for
mining)
(Done with data preprocessing)
 Data mining (an essential process where intelligent methods are applied to extract
data patterns)
 Pattern evaluation (indentify the truly interesting patterns)
 Knowledge presentation (mined knowledge is presented to the user with
visualization or representation techniques)
 Use of discovered knowledge
10

Data Mining and Business Intelligence
11
Increasing potential
to support
business decisions End User
Business
Analyst
Data
Analyst
DBA
Decision
Making
Data Presentation
Visualization Techniques
Data Mining
Information Discovery
Data Exploration
Statistical Summary, Querying, and Reporting
Data Preprocessing/Integration, Data Warehouses
Data Sources
Paper, Files, Web documents, Scientific experiments, Database Systems

A typical DM System Architecture
 Database, data warehouse, WWW or other
information
repository (store data)
 Database or data warehouse server (fetch and
combine data)
 Knowledge base (turn data into meaningful groups
according to domain knowledge)
 Data mining engine (perform mining tasks)
 Pattern evaluation module (find interesting
patterns)
 User interface (interact with the user)

A typical DM System Architecture (2)

Confluence of Multiple Disciplines
14
Data Mining
Database
Technology Statistics
Information
Science
Other
Disciplines
Visualization
Machine
Learning
• Not all “Data Mining System” performs true data mining
 machine learning system, statistical analysis (small amount of data)
 Database system (information retrieval, deductive querying…)

1.3 On What Kinds of Data?
 Database-oriented data sets and applications
 Relational database, data warehouse, transactional database
 Advanced data sets and advanced applications
 Object-Relational Databases
 Temporal Databases, Sequence Databases, Time-Series databases
 Spatial Databases and Spatiotemporal Databases
 Text databases and Multimedia databases
 Heterogeneous Databases and Legacy Databases
 Data Streams
 The World-Wide Web
15

Relational Databases
 DBMS – database management system, contains a collection of
interrelated databases
e.g. Faculty database, student database, publications database
 Each database contains a collection of tables and functions to
manage and access the data.
e.g. student_bio, student_graduation, student_parking
 Each table contains columns and rows, with columns as attributes of data and rows as
records.
 Tables can be used to represent the relationships between or among multiple tables.

Relational Databases (2) – AllElectronics store

Relational Databases (3)
 With a relational query language, e.g. SQL, we will be able to find
answers to questions such as:
 How many items were sold last year?
 Who has earned commissions higher than 10%?
 What is the total sales of last month for Dell laptops?
 When data mining is applied to relational databases, we can search for
trends or data patterns.
 Relational databases are one of the most commonly available and
rich information repositories, and thus are a major data form in our study.

Data Warehouses
 A repository of information collected from multiple sources, stored
under a unified schema, and that usually resides at a single site.
 Constructed via a process of data cleaning, data integration, data
transformation, data loading and periodic data refreshing.

Data Warehouses (2)
 Data are organized around major subjects, e.g. customer,
item, supplier and activity.
 Provide information from a historical perspective (e.g. from the
past 5 – 10 years)
 Typically summarized to a higher level (e.g. a summary of the
transactions per item type for each store)
 User can perform drill-down or roll-up operation to view the
data at different degrees of summarization

Transactional Databases
 Consists of a file where each record represents a transaction
 A transaction typically includes a unique transaction ID and a
list of the items making up the transaction.
 Either stored in a flat file or unfolded into relational tables
 Easy to identify items that are frequently sold together

1.4 Data Mining Functionalities
- What kinds of patterns can be mined?
 Concept/Class Description: Characterization and
Discrimination
 Data can be associated with classes or concepts.
 E.g. classes of items – computers, printers, …
concepts of customers – bigSpenders, budgetSpenders, …
 How to describe these items or concepts?
 Descriptions can be derived via
 Data characterization – summarizing the general characteristics of a
target class of data.
 E.g. summarizing the characteristics of customers who spend more than $1,000 a year
at AllElectronics. Result can be a general profile of the customers, such as 40 – 50 years
old, employed, have excellent credit ratings.
23

 Data discrimination – comparing the target class with one or a set of
comparative classes
 E.g. Compare the general features of software products whole sales increase by 10% in
the last year with those whose sales decrease by 30% during the same period
 Or both of the above
 Mining Frequent Patterns, Associations and
Correlations
 Frequent itemset: a set of items that frequently appear
together in a transactional data set (e.g. milk and bread)
 Frequent subsequence: a pattern that customers tend to purchase
product A, followed by a purchase of product B
24

 Association Analysis: find frequent patterns
 E.g. a sample analysis result – an association rule:
buys(X, “computer”) => buys(X, “software”) [support = 1%, confidence =
50%]
(if a customer buys a computer, there is a 50% chance that she will buy
software. 1% of all of the transactions under analysis showed that computer
and software
are purchased together. )
 Associations rules are discarded as uninteresting if they do not satisfy both a
minimum support threshold and a minimum confidence threshold.
 Correlation Analysis: additional analysis to find statistical
correlations between associated pairs
25

 Classification and Prediction
 Classification
 The process of finding a model that describes and distinguishes the data
classes or concepts, for the purpose of being able to use the model to
predict the class of
objects whose class label is unknown.
 The derived model is based on the analysis of a set of training data (data
objects whose class label is known).
 The model can be represented in classification (IF-THEN) rules, decision
trees,
neural networks, etc.
 Prediction
 Predict missing or unavailable numerical data values
26

27

Data Mining Functionalities (2)
 Cluster Analysis
 Class label is unknown: group data to form new classes
 Clusters of objects are formed based on the principle of
maximizing intra-class similarity & minimizing interclass similarity
 E.g. Identify homogeneous subpopulations of customers. These clusters may
represent individual target groups for marketing.
28

Data Mining Functionalities (2)
 Outlier Analysis
 Data that do no comply with the general behavior or model.
 Outliers are usually discarded as noise or exceptions.
 Useful for fraud detection.
 E.g. Detect purchases of extremely large amounts
 Evolution Analysis
 Describes and models regularities or trends for objects whose
behavior changes over time.
 E.g. Identify stock evolution regularities for overall stocks and for the stocks of
particular companies.
29

1.5 Are All of the Patterns Interesting?
 Data mining may generate thousands of patterns: Not all of
them
are interesting
 A pattern is interesting if it is
 easily understood by humans
 valid on new or test data with some degree of certainty,
 potentially useful
 novel
 validates some hypothesis that a user seeks to confirm
 An interesting measure represents knowledge !
30

 Objective measures
 Based on statistics and structures of patterns, e.g., support,
confidence, etc. (Rules that do not satisfy a threshold are considered
uninteresting.)
 Subjective measures
 Reflect the needs and interests of a particular user.
 E.g. A marketing manager is only interested in characteristics of customers who shop
frequently.
 Based on user’s belief in the data.
 e.g., Patterns are interesting if they are unexpected, or can be used for strategic
planning, etc
 Objective and subjective measures need to be combined.
31

 Find all the interesting patterns: Completeness
 Unrealistic and inefficient
 User-provided constraints and interestingness measures should be
used
 Search for only interesting patterns: An optimization problem
 Highly desirable
 No need to search through the generated patterns to identify truly
interesting ones.
 Measures can be used to rank the discovered patterns according
their
interestingness.
32

1.6 Classification of data mining systems
Data Mining
Database
Technology Statistics
Information
Science
Other
Disciplines
Visualization
Machine
Learning

1.6 Classification of data mining systems
 Database
 Relational, data warehouse, transactional, stream, object-oriented/relational,
active, spatial, time-series, text, multi-media, heterogeneous, legacy, WWW
 Knowledge
 Characterization, discrimination, association, classification, clustering,
trend/deviation, outlier analysis, etc.
 Multiple/integrated functions and mining at multiple levels
 Techniques utilized
 Database-oriented, data warehouse (OLAP), machine learning, statistics,
visualization, etc.
 Applications adapted
 Retail, telecommunication, banking, fraud analysis, bio-data mining, stock
market analysis, text mining, Web mining, etc.

1.7 Data Mining Task Primitives
 How to construct a data mining query?
 The primitives allow the user to interactively communicate with
the data mining system during discovery to direct the mining
process, or examine the findings
35

 The primitives specify:
(1) The set of task-relevant data – which portion of the database to be
used
 Database or data warehouse name
 Database tables or data warehouse cubes
 Condition for data selection
 Relevant attributes or dimensions
 Data grouping criteria
36

 The primitives specify:
(2) The kind of knowledge to be mined – what DB functions to be
performed
 Characterization
 Discrimination
 Association
 Classification/prediction
 Clustering
 Outlier analysis
 Other data mining tasks
37

(3) The background knowledge to be used – what domain knowledge,
concept hierarchies, etc.
(4) Interestingness measures and thresholds – support, confidence, etc.
(5) Visualization methods – what form to display the result, e.g. rules,
tables, charts, graphs, …
38

 DMQL – Data Mining Query Language
 Designed to incorporate these primitives
 Allow user to interact with DM systems
 Providing a standardized language like SQL
39

An Example Query in DMQL
40
(1)
(3)
(2)
(1)
(1)
(1)
(2)
(1)
(5)

Why Data Mining Query Language?
 Automated vs. query-driven?
 Finding all the patterns autonomously in a database?—
unrealistic because the patterns could be too many but
uninteresting
 Data mining should be an interactive process
 User directs what to be mined
 Users must be provided with a set of primitives to be used to
communicate with the data mining system
 Incorporating these primitives in a data mining query language
 More flexible user interaction
 Foundation for design of graphical user interface
 Standardization of data mining industry and practice
41

1.8 Integration of Data Mining and Data
Warehousing
42
• No coupling
– Flat file processing, no utilization of any functions of a
DB/DW
system
– Not recommended
• Loose coupling
– Fetching data from DB/DW
– Does not explore data structures and query optimization
methods provided by DB/DW system
– Difficult to achieve high scalability and good performance
with
large data sets

1.8 Integration of Data Mining and Data
Warehousing
43
• Semi-tight
– Efficient implementations of a few essential data mining
primitives in a DB/DW system are provided, e.g., sorting,
indexing, aggregation,
histogram analysis, multiway join, precomputation of some stat
functions
– Enhanced DM performance
• Tight
– DM is smoothly integrated into a DB/DW system, mining query is
optimized based on mining query analysis, data structures,
indexing, query processing methods of a DB/DW system
– A uniform information processing environment, highly desirable

1.9 Major Issues in Data Mining
 Mining methodology and User interaction
 Mining different kinds of knowledge
 DM should cover a wide spectrum of data analysis and knowledge discovery tasks
 Enable to use the database in different ways
 Require the development of numerous data mining techniques
 Interactive mining of knowledge at multiple levels of abstraction
 Difficult to know exactly what will be discovered
 Allow users to focus the search, refine data mining requests
 Incorporation of background knowledge
 Guide the discovery process
 Allow discovered patterns to be expressed in concise terms and different levels of
abstraction
 Data mining query languages and ad hoc data mining
 High-level query languages need to be developed
 Should be integrated with a DB/DW query language
44

 Presentation and visualization of results
 Knowledge should be easily understood and directly usable
 High level languages, visual representations or other expressive forms
 Require the DM system to adopt the above techniques
 Handling noisy or incomplete data
 Require data cleaning methods and data analysis methods that can
handle noise
 Pattern evaluation – the interestingness problem
 How to develop techniques to access the interestingness of discovered
patterns, especially with subjective measures bases on user beliefs or
expectations
45

 Performance Issues
 Efficiency and scalability
 Huge amount of data
 Running time must be predictable and acceptable
 Parallel, distributed and incremental mining algorithms
 Divide the data into partitions and processed in parallel
 Incorporate database updates without having to mine the entire data again from
scratch
 Diversity of Database Types
 Other database that contain complex data objects, multimedia data,
spatial data, etc.
 Expect to have different DM systems for different kinds of data
 Heterogeneous databases and global information systems
 Web mining becomes a very challenging and fast-evolving field in data mining
46

Data Mining: Concepts and Techniques.ppt

More Related Content

Similar to Data Mining: Concepts and Techniques.ppt (20)

Recently uploaded (20)

Data Mining: Concepts and Techniques.ppt