2 pages due in 5 hours
Academicauthority
4-MIST640-PredictiveAnalytics-Privacy.pptxMIST 640 Class Notes 4
Predictive Analytics & Data Mining
MIST 640
Business Intelligence and Performance Management
Contents
Data Mining
Data Mining Process
Data Mining Goes to Hollywood! Example
Video Case: Predictive Analytics in Law Enforcement
Data Mining Problems
Predictive Accuracy
Decision Trees
Video Case: Fraud Detection with Predictive Analytics
Data ownership and Privacy
Video Case: Data Mining in Retail
2
MIST 640
Business Intelligence and Performance Management
Definition of Data Mining
Data mining is a process of discovering and interpreting previously unknown patterns in data to solve business problems.
Data mining is a form of discovery-driven analysis where statistical and machine-learning techniques are used to make predictions or estimates about outcomes or traits before knowing their true values.
Data mining is an iterative process: the lessons learned during the data mining process and from the deployed solution can trigger new, often more focused business questions.
3
MIST 640
Business Intelligence and Performance Management
Data Mining Primer for the Data Warehouse Professional, Arlene Zaima, James Kashner 2003
The CRISP-DM Model: The New Blueprint for Data Mining, Colin Shearer, 2000
3
Evolution of Business Intelligence
4
MIST 640
Business Intelligence and Performance Management
4
Predictive Analytics Example: Computer Purchases
5
Would a person with the following characteristics
(age<=30, Income=medium, Student=yes, Credit_rating= Fair)
buy computer?
Unknown outcome
Known
outcomes
MIST 640
Business Intelligence and Performance Management
Example from Tan, Steinbach, Kumar 2004
Multimedia Lecture Support Package to Accompany Basic Marketing
Lecture Script 6-5
Video Example
In the movie Minority Report, there is a system, which is used to predict future murders. But, is it perfect?
http:// www.youtube.com/watch?v=BdUxwYScrFo
Advertising in Minority Report
6
MIST 640
Business Intelligence and Performance Management
Predictive vs. Descriptive Models
Predictive models are used to predict an outcome, referred to as the dependent or target variable, based on the values of other variables (independent) variables in the data set.
The likelihood that a customer will purchase a product based on her income, number of children, current product ownership or debt.
Descriptive models describe a particular pattern that has no known outcome.
Visualization: reducing data to a picture that can be easily understood.
Clustering: group data into subsets based on similar values of attributes, e.g. customer segments like loyal or high-value customer segment
7
MIST 640
Business Intelligence and Performance Management
The CRISP-DM Model: The New Blueprint for Data Mining, Colin Shearer, 2000
7
Data Mining Process: CRISP-DM
8
MIST 640
Business Intelligence and Performance Management
8
Data Mining Process: CRISP-DM
Step 1: Business Understanding
Step 2: Data Understanding
Step 3: Data Preparation (!)
Step 4: Model Building
Step 5: Testing and Evaluation
Step 6: Deployment
The process is highly repetitive and experimental (DM: art versus science?)
9
Accounts for ~85% of total project time
MIST 640
Business Intelligence and Performance Management
9
1. Business Understanding
Understanding the project objectives from a business perspective, converting this knowledge into a data mining problem definition, and then developing a preliminary plan designed to achieve the objectives.
Successful data mining begins with a clearly defined business objective.
Everything from data pre-processing to model selection is driven by the business objective.
10
MIST 640
Business Intelligence and Performance Management
The CRISP-DM Model: The New Blueprint for Data Mining, Colin Shearer, 2000
10
1. Business Objective Examples
For a bank: Retain current customers by predicting when they are prone to move to a competitor
Related questions
How does the primary channel (e.g., ATM, branch visit, Internet) of a bank customer affect whether they stay or go?
Will lower ATM fees significantly reduce the number of high value customers who leave?
How will lower ATM fees affect other segments?
Target: Reduce lost customers by 10 percent
Data mining objective: Predict how many customers will leave given their transactions in the past three years, demographic information (age, salary, city, etc.), and the ATM fee.
Target: 85% predictive accuracy
11
MIST 640
Business Intelligence and Performance Management
The CRISP-DM Model: The New Blueprint for Data Mining, Colin Shearer, 2000
Next class
11
Other Business Objectives/Questions
12
MIST 640
Business Intelligence and Performance Management
Data Mining Primer for the Data Warehouse Professional, Arlene Zaima, James Kashner 2003
12
Testing and Evaluation
Simple split (or holdout or test sample estimation)
Split the data into 2 mutually exclusive sets training (~70%) and testing (30%)
For neural networks, the data is split into three sub-sets (training [~60%], validation [~20%], testing [~20%])
13
MIST 640
Business Intelligence and Performance Management
13
Classification Model Construction
14
Training
Data
Classification
Algorithms
IF rank = ‘professor’
OR years > 6
THEN tenured = ‘yes’
Classifier
(Model)
MIST 640
Business Intelligence and Performance Management
Web source
14
Use the Classification Model in Prediction
15
Classifier
Testing
Data
Unseen Data
(Jeff, Professor, 4)
Tenured?
MIST 640
Business Intelligence and Performance Management
Web source
15
Video Case: Predictive Analytics in Law Enforcement
Links:
Introduction Predictive Analytics - Police Use Analytics to Reduce Crime http :// youtu.be/8SJQtn4RO7I
Concept Predictive Analytics: Using Real-Time Data to help transform Public… http :// youtu.be/6FEKGRf3mLo
Memphis PD case: Memphis PD: Fighting Crime with Analytics http :// youtu.be/wi7Pbm6tHws
Identify what kind of data is involved for this problem.
Identify the independent and dependent variables in this case.
Describe the prediction problem.
What is the value-added of this system to Memphis?
16
MIST 640
Business Intelligence and Performance Management
Next class may 26th
16
Video Case: Fraud Detection with Predictive Analytics
Links:
Credit Card Fraud Prediction problem http:// youtu.be/WO96ztcFVp8
Healthcare Insurance Fraud Fraud Detection and Prevention with SAS -- Health Care Payers http:// youtu.be/5EtcR3ZqbRs
Auto Insurance Fraud Infinity Insurance: Driving the auto insurance industry forward http:// youtu.be/e6GmfnCND5o
Brainstorm what kind of data may be involved for each type of fraud.
Identify the independent and dependent variables for each type of fraud.
What is the prediction problem? Characterize false positives and negatives? Which one is more serious?
What is the value-added of fraud detection in each industry?
17
MIST 640
Business Intelligence and Performance Management
False Positives and False Negatives
Confusion/classification matrix for fraud detection
18
MIST 640
Business Intelligence and Performance Management
Accuracy of Classification Models
In classification problems, the primary source for accuracy estimation is the confusion matrix (classification matrix)
19
MIST 640
Business Intelligence and Performance Management
19
Privacy
The right of privacy is the right of a person to be free from
intrusion upon a person's seclusion or solitude,
the unwarranted appropriation or exploitation of one’s personality for the use or benefit of the wrongdoer,
the publicizing of one’s private affairs with which the public has no legitimate concern,
public disclosure of embarrassing private facts about someone;
publicity which places someone in a false light in the public eye,
the wrongful intrusion into one’s private activities in such manner as to outrage or cause mental suffering, shame or humiliation to a person of ordinary sensibilities.
20
MIST 640
Business Intelligence and Performance Management
http://privacy.uslegal.com/#sthash.yPmS8MkA.dpuf
http://www.law.washington.edu/Directory/docs/Winn/Who%20Owns%20the%20Customer.htm
20
Data Ownership
Data ownership refers to both the possession of and responsibility for information.
Ownership implies power as well as control.
The control of information includes not just the ability to access, create, modify, package, derive benefit from, sell or remove data, but also the right to assign these access privileges to others.
At the core, the degree of ownership (and by corollary, the degree of responsibility) is driven by the value that each interested party derives from the use of the data.
Do consumers own data collected on themselves? Do the companies own the data? Do companies have a right to sell their customers’ data?
21
MIST 640
Business Intelligence and Performance Management
http://www.niu.edu/rcrportal/datamanagement/dotopic.html
21
Attitudes towards Privacy
The privacy Fundamentalists: Fundamentalists are generally distrustful of organizations that ask for their personal information, worried about the accuracy of computerized information and additional uses made of it.
The Pragmatic: They weigh the benefits to them of various consumer opportunities and services, protections of public safety or enforcement of personal morality against the degree of intrusiveness of personal information sought and the increase in government power involved.
The Unconcerned: The Unconcerned are generally trustful of organizations collecting their personal information, comfortable with existing organizational procedures and uses
22
MIST 640
Business Intelligence and Performance Management
Kumaraguru, P & Cranor, L. (2005). Privacy Indexes: A Survey of Westin’s Studies. http://reports-archive.adm.cs.cmu.edu/anon/anon/usr0/ftp/usr/ftp/isri2005/CMU-ISRI-05-138.pdf
22
Importance of Privacy
23
70% of people worry about the erosion of personal privacy, while 78% worry about a further global financial crisis.
MIST 640
Business Intelligence and Performance Management
McCann Truth Central. The Truth about Privacy. http://mccann.com/wp-content/uploads/2012/06/ McCann_Truth_about_Privacy.pdf Oct. 18, 2011.
23
Prime Concerns about Privacy
The security of finances
The security of reputation
24
MIST 640
Business Intelligence and Performance Management
McCann Truth Central. The Truth about Privacy. http://mccann.com/wp-content/uploads/2012/06/ McCann_Truth_about_Privacy.pdf Oct. 18, 2011.
24
Prime Concerns about Privacy
25
MIST 640
Business Intelligence and Performance Management
McCann Truth Central. The Truth about Privacy. http://mccann.com/wp-content/uploads/2012/06/ McCann_Truth_about_Privacy.pdf Oct. 18, 2011.
25
What data to share and What data to keep private The Consumers understand that there are major benefits associated with sharing data with businesses online.
26
MIST 640
Business Intelligence and Performance Management
McCann Truth Central. The Truth about Privacy. http://mccann.com/wp-content/uploads/2012/06/ McCann_Truth_about_Privacy.pdf Oct. 18, 2011.
26
Right to Privacy is not equal The more a person gains benefit from public, the less they have a right to privacy.
27
MIST 640
Business Intelligence and Performance Management
McCann Truth Central. The Truth about Privacy. http://mccann.com/wp-content/uploads/2012/06/ McCann_Truth_about_Privacy.pdf Oct. 18, 2011.
27
Trust depends on the company (2011)
28
MIST 640
Business Intelligence and Performance Management
McCann Truth Central. The Truth about Privacy. http://mccann.com/wp-content/uploads/2012/06/ McCann_Truth_about_Privacy.pdf Oct. 18, 2011.
28
Assurance: I know exactly how my data is going to be used.
Commitment: The company doesn’t pass my data on to any third parties.
Choice: I can control exactly which pieces of information I share and don’t share.
Value: I have a clear understanding of how giving up my data will benefit me, i.e. what’s in it for me.
Keys to Trust
29
MIST 640
Business Intelligence and Performance Management
McCann Truth Central. The Truth about Privacy. http://mccann.com/wp-content/uploads/2012/06/ McCann_Truth_about_Privacy.pdf Oct. 18, 2011.
29
Video Case: Data Mining in Retail
Links:
DATA MINING | The Checkout http:// www.youtube.com/watch?v=f2Kji24833Y
Questions
What is the inference problem?
What is the value created by data mining?
For the business
For the consumer
For the society
Is privacy an issue of give-and-take, i.e. something for something? Is there fair sharing of value?
What should be the role of prediction in society?
30
MIST 640
Business Intelligence and Performance Management
ageincomestudentcredit_ratingbuys_computer
<=30highnofairno
<=30highnoexcellentno
30…40highnofairyes
>40mediumnofairyes
>40lowyesfairyes
>40lowyesexcellentno
31…40lowyesexcellentyes
<=30mediumnofairno
<=30lowyesfairyes
>40mediumyesfairyes
<=30mediumyesexcellentyes
31…40mediumnoexcellentyes
31…40highyesfairyes
>40mediumnoexcellentno
Sheet1
| age | income | student | credit_rating | buys_computer |
| <=30 | high | no | fair | no |
| <=30 | high | no | excellent | no |
| 30…40 | high | no | fair | yes |
| >40 | medium | no | fair | yes |
| >40 | low | yes | fair | yes |
| >40 | low | yes | excellent | no |
| 31…40 | low | yes | excellent | yes |
| <=30 | medium | no | fair | no |
| <=30 | low | yes | fair | yes |
| >40 | medium | yes | fair | yes |
| <=30 | medium | yes | excellent | yes |
| 31…40 | medium | no | excellent | yes |
| 31…40 | high | yes | fair | yes |
| >40 | medium | no | excellent | no |
Data Sources
Business
Understanding
Data
Preparation
Model
Building
Testing and
Evaluation
Deployment
Data
Understanding
6
12
3
5
4
Preprocessed
Data
Training Data
Testing Data
Model
Development
Model
Assessment
(scoring)
2/3
1/3
Classifier
Prediction
Accuracy
NAMERANKYEARSTENURED
MikeAssistant Prof3no
MaryAssistant Prof7yes
Bill Professor2yes
JimAssociate Prof7yes
DaveAssistant Prof6no
AnneAssociate Prof3no
Sheet1
| NAME | RANK | YEARS | TENURED |
| Mike | Assistant Prof | 3 | no |
| Mary | Assistant Prof | 7 | yes |
| Bill | Professor | 2 | yes |
| Jim | Associate Prof | 7 | yes |
| Dave | Assistant Prof | 6 | no |
| Anne | Associate Prof | 3 | no |
NAME
RANK
YEARS
TENURED
Tom
Assistant Prof
2
no
Merlisa
Associate Prof
7
no
George
Professor
5
yes
Joseph
Assistant Prof
7
yes
Sheet1
| NAME | RANK | YEARS | TENURED |
| Tom | Assistant Prof | 2 | no |
| Merlisa | Associate Prof | 7 | no |
| George | Professor | 5 | yes |
| Joseph | Assistant Prof | 7 | yes |
FN
TP
TP
Rate
Positive
True
+
=
FP
TN
TN
Rate
Negative
True
+
=
FN
FP
TN
TP
TN
TP
Accuracy
+
+
+
+
=
FP
TP
TP
recision
+
=
P
FN
TP
TP
call
Re
+
=
True
Positive
Count (TP)
False
Positive
Count (FP)
True
Negative
Count (TN)
False
Negative
Count (FN)
True Class
PositiveNegative
P
o
s
i
t
i
v
e
N
e
g
a
t
i
v
e
P
r
e
d
i
c
t
e
d
C
l
a
s
s
