Transcript modeling response to direct mail marketing

MODELING
RESPONSE TO DIRECT MAIL
MARKETING
ISQS 7342 – Dr. Zhangxi Lin
by
Junil Chang
Cross-Industry Standard Process for
Data Mining
• According to CRISP–DM, a given data mining
project has a life cycle consisting of six phases
Six phases for data mining
1. Business understanding phase
2. Dada understanding phase
3. Data preparation phase
4. Modeling phase
5. Evaluation phase
6. Deployment phase
Direct Mail Marketing Response
Problem
• This project is to predict which customers are
most likely to respond to a direct mail
marketing promotion.
• The clothing-store data set is provided by a
clothing store chain in New England.
• Data were collected on 33 fields for 21,740
customers
Building the Cost/Benefit Table
• Classification models are evaluated on
accuracy rates, error rates, false negative
rates, and false positive rates.
• four possible decision outcomes (true
negative, true positive, false negative, and
false positive) and assign reasonable costs to
each decision outcome.
Statistics associated with the
average amount spent per visit for all
customers
Statistics associated with the
average amount spent per visit for all
customers
• Profit: $113.59 * .25 = $28.40
• Benefit: $28.40 - $2.00 = $26.40.
Cost/Benefit Decision Summary for the
Clothing Store Marketing Promotion Problem
Outcome
Classification
Actual Response
Cost
True negative (TN)
Nonresponse
Nonresponse
No cost
True positive (TP)
Response
Response
$26.40 profit
False negative (FN)
Nonresponse
Response
$28.40 lost profit
False positive (FP)
Response
Nonresponse
$2 for the mailing cost
Clothing Store Data Set
• The original clothing-store data set contains
28799 customers in 51 fields
• Pre processed data set contains 21740
customers in 33 fields
• Distribution of response variable
• Most people do not response to the promotion
• 3611 of 21740 customers or 16.61% responses
for marketing campaign (1 indicates response,
and 0 indicates non responses).
• Question