W08 Tutorial Exercise
docx
keyboard_arrow_up
School
Simon Fraser University *
*We aren’t endorsed by this school
Course
445
Subject
Information Systems
Date
Dec 6, 2023
Type
docx
Pages
2
Uploaded by huynhnhut28102000
W08 Tutorial Exercise
As you work through the tutorial in class, insert your responses to the following questions in this word
document. Please then submit one copy per team through Canvas.
Name__________________________________ Student Number___________________
Name__________________________________ Student Number___________________
------------------------------------------------------------------------------------------------------------------------------------
This has slightly more detailed instructions than Exercise 8.3.5 in
the text. Answer this one!
Exercise 8.3.5.
Run the two-node neural network model with
all
of the
remaining variables used as predictors. Use the three
newly
computed
variables (Log.LastDonAmt, Log.AveDonAmt, and New.Region) as in the
previous models
, not the original
variables. Make sure you do not enter the
target, MonthGive, or Sample into the predictor list.
a)
Compare your new model with the MixedCCS2 model by generating lift
charts on both the estimation and validation data sets. Copy them into
this Word document. Briefly describe your results, comparing the
estimation and validation lifts.
The lift chart shows a good lift. Both lines meet at the 60% of sample proportion, which means we can
have about 90% responses from the targets with only targeting 60% of the total donors. On the
Estimation Lift Chart the NNHL2New does a better result while on the Validation Lift Chart the
MixedCCS2 does a better result.
Because on the Validation Lift Chart, the logistic regression (MixedCCS2) does a better result, with
approximately 40 percentile level. It shows that there is no further nonlinear effects for the neural
network to pick up.
b)
Also compare your validation lift chart with the MixedCCS2 and NNHL2
lift in Figure 8.9.
Do you know anything more about the relations in
the data that you did not know before generating these last lift charts?
The NNHL2.New seems to have a lower percentile level compared to the
NNHL2 in Figure 8.9. Also, in the Figure 8.9, the two lines meet at about 40%
of Sample Proportion, while they meet at around 60% in the validation lift
chart. It means there is a chance to capture the same percentage of
responses from targeting less sample, suggested from the lift chart of Figure
8.9. Therefore, the first five predictors suggested a better result.
c)
A question to challenge your understanding and creativity: Do these
results suggest any next steps? (Hint: Can we sell a black box to
managers?)
Your preview ends here
Eager to read complete document? Join bartleby learn and gain access to the full version
- Access to all documents
- Unlimited textbook solutions
- 24/7 expert homework help