Segmentation Applications in Banking - Data Mining Techniques in CRM: Inside Customer Segmentation - page 236

Table 6.4 ( continued )

Rotated component matrix

Components

12345 6 7 8

PRC ENTERTAINMENT − 0.36 0.31 0.55

PRC EDUCATION 0.51

PRC CULTURE 0.50

PRC BOOKS MUSIC 0.36

Trivial correlations (below 0.3 in absolute value) are omitted. As a reminder,
we note that values close to 1, in absolute value, signify strong correlations and
original fields that are associated with a specific component. The interpretation
and labeling of the derived components is summarized in Table 6.5.

Table 6.5 Interpretation of the extracted principal components.

Derived Components

Component Label and Description

1. Fashion purchases:

Component 1 represents fashion purchases. Purchase of clothes and
accessories are correlated with each other and with component 1.
Therefore customers with high spending at these merchants will
also have a high (positive) component 1 score

2. Daily needs spending:

Component 2 seems to measure the relative card usage for purchases
at supermarkets and gas stations

3. Appliance purchases:

The next component is associated with purchases of appliances.
Travel spending is also (moderately) negatively correlated with
this component and with appliance purchases: not many appliance
buyers spent money on traveling. Hence this component captures
this contrast: travelers are expected to have a relatively high positive
score on component 3, whereas appliance buyers have a large
negative score

4. Spending on telecommunication services:

Component 4 is associated with using the card for telecommunication
services. Large negative scores suggest increased relevant usage