Direct Mailing to Airline Customers. East-West Airlines has entered into a partnership

with the wireless phone company Telcon to sell the latter’s service via direct

mail. The file contains a subset of a data sample of

who has already received a test offer. About 13% accepted.

You are asked to develop a model to classify East-West customers as to whether they

purchased a wireless phone service contract (target variable Phone..sale), a model

that can be used to predict classifications for additional customers.

a. Create a validation column (stratified on Phone_Sale). Then run a neural net model

on these data. Request lift curves, and interpret the meaning (in business terms)

of the lift curve for the validation set.

b. Comment on the difference between the training and validation lift curves.

c. Run a second neural net model on the data, this time setting the number of tours

to 20. Comment now on the difference between this model and the model you ran

earlier.

d. Run a third neural net model on the data, with one tour. This time add a second

hidden layer, different activation functions, and several nodes. Comment on the

difference between this model and the first model you ran, and how overfitting

might have affected results.

e. What sort of information, if any, is provided about the effects and importance of

the various variables?

f. For this assignment, we did not ask you to set the random seed when creating the

validation column or when building the models.

i. Comment on why we might set the random seed before creating the validation

column.

ii. Comment on why we might set the random seed before building each neural

network model.