The parameter iApplyOutputPrefix is the prefix of the table used to
contain cluster assignments. For example, if iApplyOutPrefix is set to
“Clusters,” and iModelPrefix is set to “MyCluster,” the system will
apply the clustering models to generate a table called “Clusters5” in
which there will be an attribute for the customer identifier and an
attribute called “MyCluster5” containing the clusters associated with
In fact, we have left the possibility for the user, through the Bool-
ean member variable mUseApplyOutPrefix , to either create a new
table for each clustering model (this is the default option that is
described here) or to create a single table (in this case the prefix will
not be concatenated with the number of clusters) with one column
for each clustering model. In general, the second solution requires
less database maintenance because fewer tables are created, but the
applications of the different clustering models will require either
inserting or updating in this table. The first clustering model will
insert the lines for the first assignment between customer identifiers
and clusters, but all the other models will have to update the values
for the new clustering assignments and this update operation can be
long if the number of customers is large.
Figure 12-4 shows the different situations when models with 5 to 7
clusters are built.
Tables containing cluster association using different settings.