Segmentation Applications in Banking - Data Mining Techniques in CRM: Inside Customer Segmentation - page 233

Table 6.2 ( continued )

Field name Description

PRC_FULL_PAYMENT Percentage of months with full payment of the due statement

balance

PRC_ZERO_PAYMENT Percentage of months with no payments at all of the due

statement balance

LIMIT_USAGE The ratio of average balance to credit limit.

PRC_INTERNATIONAL Amount spent abroad as a percentage of total spending

amount

PRC_INTERNET Web purchases as a percentage of total purchase amount

Credit account information fields - profiling fields

Customer socio-demographics - profiling fields

THE ANALYTICAL PROCESS

The analytical process that was followed included three main steps.

1. Extraction of the segmentation dimensions by using a supervised data reduction
technique.
2. Identification of the customer segments through the application of a clustering
model.
3. Profiling of the segments.

In Figure 6.2 we can see the IBM SPSS Modeler stream developed for the
needs of segmentation. The data preparation part is omitted in this screenshot.
The first step involved the application of a PCA command (node) for the data
reduction and identification of the underlying data dimensions. Customers were
then scored by the generated PCA model (the ''diamond'' node) and, in the final
step of the procedure, the generated PCA scores were loaded as inputs for training
a TwoStep cluster model.
The procedure concluded with the exhaustive profiling and evaluation of
the derived solution, before accepting it and adopting it into the organization's
procedures. The whole process is presented in detail in the following sections.

Revealing the Segmentation Dimensions

The clustering inputs included the set of fields denoting the percentage of
purchases per merchant category. These fields were used instead of the average