Databases Reference
In-Depth Information
sports equipment by each of the sampled Big Box customers. There are 5,000
customers in the sample.
Big Box would like to direct some targeted magazine based promotions at
customers most likely to respond. To assist in their analysis, Big Box surveyed
each of these 5,000 customers to determine which magazines they read. Their
responses are in the dataset MagazineReadership.csv. For each magazine
surveyed (Cosmopolitan, Jet, Newsweek, PCMag, People, and Sports Illus-
trated), readership is coded as 1 and non-readership as 0. There is no in-between
“occasionally” option.
Join the two datasets by dragging MagazineReadership.csv and dropping it
on OutdoorCustomerSales.csv; then select “Join Datasets”.
In the Join Parameters dialog, check CustomerID as the matching column
in each of the datasets.
Click “OK”.
Create a derived dataset named “data” from the resulting joined dataset.
Include the following columns in the dataset: Cosmopolitan, Jet, News-
week, PCMag, People, SportsIllustrated, and Purchases.
Create a derived dataset named “readership” from “data” to be used for the
cluster analysis. It should include all the six magazine columns, but not
Purchases.
Create a 3
3
3 clustering of readership.
View the SOM:readership model in the SOM viewer.
Open “data” in a parallel plot synchronized with the SOM:readership
clustering.
Drag the Purchases column all the way to the right to keep it separated from
the six magazine readership columns.
Check “Show Means”.
Click on the largest cluster.
The largest cluster contains 2,434 observations. In the parallel plot, notice
that readership is zero for all six magazines. This is the cluster representing
customers that read none of the six surveyed magazines.
