Download An Introduction to Data Mining

An Introduction to Data Mining
g
BY:GAGAN DEEP
KAUSHAL
Trends leading
g to Data Flood
ƒ More data is generated:
ƒ Bank, telecom, other business
transactions ...
ƒ Scientific Data: astronomy,
biology, etc
ƒ Web, text, and e-commerce
ƒ More data is captured:
ƒ Storage technology faster and
cheaper
ƒ DBMS capable of handling
COM 307: Machine Learning and Data
bigger DB
Mining
2
Growth Trends
ƒ Moore’s law
ƒ Computer Speed doubles every 18
months
ƒ Storage law
ƒ total storage doubles every 9 months
ƒ Consequence
ƒ very little data will ever be looked at by
a human
h
ƒ Data Mining is NEEDED to make
sense and use of data
data.
COM 307: Machine Learning and Data
Mining
3
Data Mining Definition
Data mining in Data is the
non-trivial process of identifying
ƒ valid
ƒ novel
ƒ potentially useful
ƒ and ultimately understandable patterns in
data.
COM 307: Machine Learning and Data
Mining
4
What is Data Mining?
z Many
Definitions
– Non-trivial
o t a e
extraction
t act o o
of implicit,
p c t, p
previously
e ous y unknown
u
o
and potentially useful information from data
– Exploration & analysis, by automatic or
semi-automatic means, of
large quantities of data
in order to discover
meaningful patterns
What is (not) Data Mining?
What is not Data
Mining?
z
z
What is Data Mining?
– Look up phone
number in phone
directory
– Certain names are more
prevalent in certain US
S
locations (O’Brien, O’Rurke,
O’Reilly
O
Reilly… in Boston area)
– Query a Web
search engine for
f
information about
Amazon
“Amazon”
– Group together similar
documents returned by search
engine according to their
context (e
(e.g.
g Amazon
rainforest, Amazon.com,)
z
Why Mine Data? Commercial
Vi
Viewpoint
i t
Lots of data is being collected
and warehoused
– Web data, e-commerce
– purchases at department/
grocery stores
– Bank/Credit Card
t
transactions
ti
z
Computers
p
have become cheaper
p and more p
powerful
z
Competitive Pressure is Strong
– Provide better
better, customized services for an edge (e.g.
(e g in
Customer Relationship Management)
Why Mine Data? Scientific Viewpoint
z
Data collected and stored at
enormous speeds (GB/hour)
– remote sensors on a satellite
– telescopes scanning the skies
– microarrays generating gene
expression data
– scientific simulations
generating terabytes of data
z
z
Traditional techniques infeasible for raw data
D t mining
Data
i i may h
help
l scientists
i ti t
– in classifying and segmenting data
i H
th i F
ti
Related Fields
Machine
Learning
Visualization
Data Mining and
Knowledge Discovery
Statistics
Databases
COM 307: Machine Learning and Data
Mining
9
Data Mining Process
Integration
Interpretation
& Evaluation
E l ti
Knowledge
Knowledge
__ __ __
__ __ __
__ __ __
DATA
Ware
house
Target
Data
Transformed
Data
COM 307: Machine Learning and Data
Mining
Patterns
and
Rules
Understa
anding
Raw
Data
10
Major Data Mining Tasks
ƒ Classification: predicting an item class
ƒ Associations:
A
i ti
e.g. A & B & C occur ffrequently
tl
ƒ Visualization: to facilitate human discovery
ƒ Estimation: predicting a continuous value
ƒ Deviation Detection: finding changes
ƒ Link Analysis: finding relationships
ƒ…
COM 307: Machine Learning and Data
Mining
11
Classification: Definition
z
Given a collection of records (training set )
–E
Each
h record
d contains
t i a sett off attributes,
tt ib t
one off the
th
attributes is the class.
Find a model for class attribute as a
function of the values of other attributes.
z Goal: previously unseen records should be
assigned a class as accurately as possible.
z
– A test set is used to determine the accuracy of the
model. Usually, the given data set is divided into
training and test sets, with training set used to build the
model
d l and
d ttestt sett used
d tto validate
lid t itit.
Classification Example
Tid
R efund
Refund
M arital
S tatus
Taxable
Incom e
C heat
Marital
Status
Taxable
Incom e
Cheat
1
Y es
S ingle
125K
No
No
Single
75K
?
2
No
M arried
100K
No
Yes
Married
50K
?
3
No
S ingle
70K
No
No
Married
150K
?
4
Y es
M arried
120K
No
Yes
Divorced
90K
?
5
No
D ivorced
i
d
95K
Y es
N
No
Si l
Single
40K
?
6
No
M arried
60K
No
No
Married
80K
?
10
7
Y es
D ivorced
220K
No
8
No
S ingle
85K
Y es
9
No
M arried
75K
No
10
10
No
S ingle
90K
Y es
Training
Set
Learn
Classifier
Test
Set
Model
Classification: Application 1
z
Direct Marketing
– Goal:
G l Reduce
R d
costt off mailing
ili b
by targeting
t
ti a sett off
consumers likely to buy a new cell-phone product.
– Approach:
‹
Use the data for a similar product introduced before.
‹
We know which customers decided to buy and which decided
otherwise. This {buy, don’t buy} decision forms the class attribute.
‹
Collect various demographic, lifestyle, and company-interaction
related information about all such customers.
– Type of business, where they stay, how much they earn, etc.
‹
Use this information as input attributes to learn a classifier model.
From [Berry & Linoff] Data Mining Techniques, 1997
Classification: Application 2
z
Fraud Detection
– Goal: Predict fraudulent cases in credit card transactions
transactions.
– Approach:
‹
Use credit card transactions and the information on its accountholder as attributes.
– When does a customer buy, what does he buy, how often he pays
on time, etc
Label past transactions as fraud or fair transactions. This forms
the class attribute.
‹ Learn a model for the class of the transactions.
‹ Use this model to detect fraud by observing credit card
transactions on an account.
‹
Association Rule Discovery:
D fi iti
Definition
z Given a set of records each of which contain some number of
items from a g
given collection;
– Produce dependency rules which will predict occurrence
of an item based on occurrences of other items.
TID
Item s
1
2
3
4
5
Bread, C oke, M ilk
B
Beer,
B
Bread
d
Beer, C oke, D iaper, M ilk
Beer, Bread, D iaper, M ilk
C oke,
k D iiaper, M ilk
Rules Discovered:
{Milk} --> {Coke}
{Diaper, Milk} --> {Beer}
Association Rule Discovery: Application 1
z
Marketing and Sales Promotion:
– Let
L t the
th rule
l di
discovered
db
be
{Patty, … } --> {Potato Chips}
– Potato Chips as consequent => Can be used to determine
what should be done to boost its sales.
– Patties in the antecedent => Can be used to see which
products would be affected if the store discontinues
selling patties.
– Patty in antecedent and Potato chips in consequent =>
Can be used to see what products should be sold with
patty to promote sale of Potato chips!
Association Rule Discovery: Application 1
z
Supermarket shelf management.
– Goal:
G l To
T identify
id tif items
it
that
th t are bought
b
ht ttogether
th by
b
sufficiently many customers.
– Approach:
A
h P
Process th
the point-of-sale
i t f l d
data
t
collected with barcode scanners to find
dependencies among items
items.
– A classic rule -‹ If
a customer buys diaper and milk, then he is very
likely to buy beer.
Association Rule Discovery: Application 2
z
Inventory Management:
– Goal: A consumer appliance repair company wants to
anticipate the nature of repairs on its consumer products
and keep the service vehicles equipped with right parts to
reduce
d
on number
b off visits
i it tto consumer h
households.
h ld
– Approach: Process the data on tools and parts required in
previous repairs at different consumer locations and
discover the co-occurrence patterns.
z
Sequential Pattern Discovery:
D
Definition
fi
iti
Given is a set of objects, with each object associated with its own timeline of
events, find rules that predict strong sequential dependencies among different
events.
events
((A B))
z
(C)
( )
(D
(
E))
Rules are formed by first disovering patterns. Event occurrences in the patterns
are go
governed
erned b
by timing constraints
constraints.
(A B)
(C)
(D E)
z
Sequential Pattern Discovery:
Elogs, l
Examples
In telecommunications alarm
– (Inverter_Problem Excessive_Line_Current)
(Rectifier_Alarm) --> (Fire_Alarm)
z
In point-of-sale transaction sequences,
– Computer
C
t Bookstore:
B k t
(Intro_To_Visual_C) (C++_Primer) -->
(Perl_for_dummies,Tcl_Tk)
– Athletic Apparel Store:
(Shoes) (Racket,
(Racket Racketball) -->
> (Sports_Jacket)
(Sports Jacket)
Regression
z
z
z
Predict a value of a given continuous valued variable based
on the values of other variables, assuming a linear or
nonlinear model of dependency.
Greatly studied in statistics, neural network fields.
Examples:
– Predicting sales amounts of new product based on
advetising expenditure.
– Predicting wind velocities as a function of temperature,
humidity air pressure
humidity,
pressure, etc
etc.
– Time series prediction of stock market indices.
Deviation/Anomaly Detection
z Detect
significant deviations from normal behavior
z Applications:
– Credit Card Fraud Detection
– Network Intrusion
Detection
Typical network traffic at University level may reach over 100 million connections per day
Challenges of Data Mining
Scalability
z Dimensionality
z Co
Complex
pe a
and
d Heterogeneous
e e oge eous Data
aa
z Data Quality
z Data
D t O
Ownership
hi and
d Di
Distribution
t ib ti
z Privacy Preservation
z Streaming Data
z
●
.
THANKS
FOR
YOUR
KIND
ATTENTION