Databases Reference
In-Depth Information
Right-click on the large cluster; select “Rename cell”. Give it a more
meaningful name: “non-readers”.
To see the names, check “Show Names”.
One by one, click on each of the other clusters. Give them a name, based on
the magazine readership that you see in the synchronized parallel plot. For
example, the cluster in the opposite corner shows high readership for
Newsweek and People. Name it “NWPeople”. Use the following names for
the other columns:
People - the cluster with 820 observations.
PCMag - with 132 observations.
Jet - the smallest cluster with 76 observations.
Cosmopolitan - with 123 observations.
Newsweek - with 324 observations.
PeopleSI - with 218 observations.
SI - with 464 observations.
Use the “Ctrl-click” option on each cluster to simultaneously select all nine
clusters.
Right-click on any cluster; select “Make dataset from all selected”.
When prompted to select columns, leave all checked. Press “OK”.
In the Control Center, view the summary statistics for the newly created
“selected” dataset.
The “selected” dataset contains all of the original dataset columns plus a
new column - ClusterName. This columnisautomaticallycreatedbythe
“Make dataset from all selected” option in the SOM viewer when multiple
cells are extracted. The cardinality of this column is 9 - one for each
selected cell. The number of rows in the dataset is still 5,000 - the original
sample size.
To summarize, at this point we have clustered the customers according to
their magazine readership, then added a column, ClusterName, to each cus-
tomer row with the name of the cluster to which it was assigned.
Right-click on “selected”; then select “Aggregate rows”.
