Lakhasly

o 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.Advanced algorithms, multiprocessor computers, massive databases Pilot, Lockheed, IBM, SGI, numerous startups (nascent industry) Prospective, proactive information delivery DATA MINING: THEORY AND PRACTICE 11 Potential Applications o Data analysis and decision support
• Market analysis and management o Target marketing, customer relationship management (CRM), market basket analysis, cross selling, market segmentation
• Risk analysis and management o Forecasting, customer retention, improved underwriting, quality control, competitive analysis
• Fraud detection and detection money laundering.On-line analytic processing (OLAP), multidimensional databases, data warehouses Pilot, Comshare, Arbor, Cognos, Microstrategy Retrospective, dynamic data delivery at multiple levels Data Mining (Emerging Today) "What's likely to happen to Boston unit sales next month? Why?"Relational databases (RDBMS), Structured Query Language (SQL), ODBC Oracle, Sybase, Informix, IBM, Microsoft Retrospective, dynamic data delivery at record level Data Warehousing & Decision Support (1990s) "What were unit sales in New England last March? Drill down to Boston."- Model selection criteria 9 Sanad AL-Maskari-Data Mining Database Vs Data Mining o Query
• Well defined
• SQL o Data
• Operational data o Output
• Precise
• Subset of database ?o Origins of Data Mining
• Statistics: more theory-based, focused on testing hypotheses
• Database system: information retrieval, deductive query answering
• Data Mining: focuses on the algorithms to extract patterns from data 3 Why Data Mining?8 8 Sanad AL-Maskari-Data Mining Data Mining Algorithm o Purpose is to fit data to a model
• Descriptive (Clustering, association, summarization, etc )
• Predictive (Classification, regression , etc.) o Model selection
• Which model is the best ?- Knowledge discovery through data mining (rules, regularities, patterns, constraints) 7 7 Sanad AL-Maskari-Data Mining Potential Applications o Data analysis and decision support
• Market analysis and management o Target marketing, customer relationship management (CRM), market basket analysis, cross selling, market segmentation.o Other Applications
• Text mining (news group, email, documents) and Web mining
• Medical applications (Cancer detection)
• DNA and bio-data analysis 12 Sanad AL-Maskari-Data Mining Market Basket Analysis Example Bread, Coke, Milk Juice, Coke, Diapers, Milk Juice, Bread, Milk Juice, Bread, Diapers, Milk Anything interesting?Diapers, Milk Bread Milk (100%) Diapers Juice(66%) Diapers Milk (100%) Customers who buy diapers also tend to buy Juice Indentify potential cross-selling opportunities among related items 13 Sanad AL-Maskari-Data Mining Bank Loan Risk Example salary loan k IF salary<k THEN not repaid not repaid repaid Given: a loan application Problem: predict if the bank should approve the loan Data: records from other loans 14 Sanad AL-Maskari-Data Mining Bank Loan Risk 1.(e.g sensors)
• Data pre-processing o Outlier detection & removal o Scaling, encoding, & feature selection 19 Data Mining Process
• Feature extraction & selection o Which features will be used in our model ?o Alternative names
• Knowledge discovery (mining) in databases (KDD), knowledge extraction, data/pattern analysis, data archeology, data dredging, information harvesting, business intelligence, etc.- Science: huge data are generated in scientific experiments which requires fast and robust analysis methods other than traditional statistical analysis.22 An Example: Relational database, transactional database, and multimedia database ID Name Age Salary($) 1 Ahmad 21 48k 2 Ali 38 72k 3 Aisha 27 52k ... Trans.Satellite images is an example of spatial data o Spatiotemporal Data
• Consist of space (distance, direction, shape , etc) and time (occurrence time, duration, etc) attributes.(I)
• Huge data volumes: web data, e-commence, purchases at supermarkets, bank/credit card transactions, sensors on a satellite, gene data, etc.29 Sanad AL-Maskari-Data Mining Data Mining Functionalities (I) o Classes/Concept description:
• Summarizing object and data accurately and precisely can be very beneficial.- Terabyte personnel hard disk is becoming more common (A terabyte is equivalent to about 2 million books) 4 4 Sanad AL-Maskari-Data Mining Why Data Mining?- Risk analysis and management o Forecasting, customer retention, improved underwriting, quality control, competitive analysis.Not a subset of database (C)Prentice Hall 10 Sanad AL-Maskari-Data Mining Why Data Mining?- Features summarizations ( e.g what are the fetures for customers who spend $5000 per year for electronics),
• Discrimination(e.g Compare between group of customers) find discriminative features between classes.(II) o Data Exponential Growth
• Industry: how to enhance current services and provide customizable solutions to customers.o Other Applications
• Text mining (news group, email, documents) and Web mining.30 Data Mining Functionalities (II) o Association: find and search for frequent item set
• Diaper ?5 5 Sanad AL-Maskari-Data Mining Data Growth We can't keep up with exponential data growth!!Credit score 800 Person 3 15 Sanad AL-Maskari-Data Mining An Example: Credit Card Fraud Detection compare Very different!Daily Transactions A profile of legitimate transactions User info.- Facebook can generate 4 million posts every minute.- More than 50 billions photos have been uploaded to Instagram .- 800 Tweets generate per second in twitter.o In 24 hours:
• 8 terabytes of data generated by Twitter.- Google handles over 5.6 billion searches per day.6 Sanad AL-Maskari-Data Mining Why Data Mining?(I) o In a minutes:
• A jet engine can generate up to 10 terabytes of data per 30 minutes.- Amazon records on average 6 million sales transactions.- Fraud detection and detection money laundering.- Medical applications (Cancer detection).- DNA and bio-data analysis.Poorly defined ?No precise query language ?Potentially fraudulent!- Exponential data growth.o Solution: Data mining!Query ?Not operational data ?Output ?Fuzzy ?5 bankruptcies 2.Unemployed for 8 years 3.0 bankruptcies 2.2 bankruptcies 2.- AT&T records 275 million phone calls.Data Explosion!Data ?Credit score 25 1.Working for the last 10 years 3.Credit score 500 Person 1 Person 2 1.Working for the last 15 years

• 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.
• Alternative names
– Knowledge discovery (mining) in databases (KDD), knowledge
extraction, data/pattern analysis, data archeology, data
dredging, information harvesting, business intelligence, etc.
• Origins of Data Mining
– Statistics: more theory-based, focused on testing hypotheses
– Database system: information retrieval, deductive query
answering
– Data Mining: focuses on the algorithms to extract patterns from
data
3
Why Data Mining? (I)
– Huge data volumes: web data, e-commence, purchases at
supermarkets, bank/credit card transactions, sensors on a satellite,
gene data, etc.
– Exponential data growth.
– Terabyte personnel hard disk is becoming more common
(A terabyte is equivalent to about 2 million books)
4
4 Sanad AL-Maskari-Data Mining
Why Data Mining? (I)
• In a minutes:
– A jet engine can generate up to 10 terabytes of data per 30 minutes.
– Facebook can generate 4 million posts every minute.
– More than 50 billions photos have been uploaded to Instagram .
– 800 Tweets generate per second in twitter.
• In 24 hours:
– 8 terabytes of data generated by Twitter.
– AT&T records 275 million phone calls.
– Google handles over 5.6 billion searches per day.
– Amazon records on average 6 million sales transactions.
5
5 Sanad AL-Maskari-Data Mining
Data Growth
We can’t keep up with exponential data growth!!
Data Explosion!
6 Sanad AL-Maskari-Data Mining
Why Data Mining? (II)
• Data Exponential Growth
– Industry: how to enhance current services and provide customizable
solutions to customers.
– Science: huge data are generated in scientific experiments which
requires fast and robust analysis methods other than traditional
statistical analysis.
• Solution: Data mining!
– Knowledge discovery through data mining (rules, regularities,
patterns, constraints)
7
7 Sanad AL-Maskari-Data Mining
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 money laundering.
• Other Applications
– Text mining (news group, email, documents) and Web mining.
– Medical applications (Cancer detection).
– DNA and bio-data analysis.
8
8 Sanad AL-Maskari-Data Mining
Data Mining Algorithm
• Purpose is to fit data to a model
– Descriptive (Clustering, association, summarization, etc )
– Predictive (Classification, regression , etc.)
• Model selection
– Which model is the best ?
– Model selection criteria
9 Sanad AL-Maskari-Data Mining
Database Vs Data Mining
• Query
– Well defined
– SQL
• Data
– Operational data
• Output
– Precise
– Subset of database
 Query
 Poorly defined
 No precise query language
 Data
 Not operational data
 Output
 Fuzzy
 Not a subset of database
©Prentice Hall
10 Sanad AL-Maskari-Data Mining
Why Data Mining? (III)
Evolutionary Step Business Question Enabling Technologies Product Providers Characteristics
Data Collection
(1960s)
"What was my total revenue in
the last five years?"
Computers, tapes, disks IBM, CDC Retrospective,
static data
delivery
Data Access
(1980s)
"What were unit sales in New
England last March?"
Relational databases
(RDBMS), Structured
Query Language (SQL),
ODBC
Oracle, Sybase,
Informix, IBM,
Microsoft
Retrospective,
dynamic data
delivery at
record level
Data Warehousing &
Decision Support
(1990s)
"What were unit sales in New
England last March? Drill down
to Boston."
On-line analytic processing
(OLAP), multidimensional
databases, data
warehouses
Pilot, Comshare,
Arbor, Cognos,
Microstrategy
Retrospective,
dynamic data
delivery at
multiple levels
Data Mining
(Emerging Today)
"What’s likely to happen to
Boston unit sales next month?
Why?"
Advanced algorithms,
multiprocessor computers,
massive databases
Pilot, Lockheed,
IBM, SGI,
numerous startups
(nascent industry)
Prospective,
proactive
information
delivery
DATA MINING: THEORY AND PRACTICE
11
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 money laundering.
• Other Applications
– Text mining (news group, email, documents) and Web mining
– Medical applications (Cancer detection)
– DNA and bio-data analysis
12 Sanad AL-Maskari-Data Mining
Market Basket Analysis Example
Bread,
Coke,
Milk
Juice, Coke,
Diapers, Milk
Juice, Bread,
Milk
Juice, Bread,
Diapers, Milk
Anything interesting?
Diapers, Milk
Bread Milk (100%)
Diapers Juice(66%)
Diapers Milk (100%)
Customers who buy diapers also tend to buy Juice
Indentify potential cross-selling opportunities among related items
13 Sanad AL-Maskari-Data Mining
Bank Loan Risk Example
salary
loan
k
IF salary