Data Mining Techniques for Segmentation - Data Mining Techniques in CRM: Inside Customer Segmentation - page 94

2.0

1.5

1.0

0.5

0.0

0.0 0.5 1.0 1.5 2.0

$KX-Kohonen

Figure 3.10 Sample scatterplot of a SOM grid map.

Table 3.11 IBM SPSS Modeler recommended Kohonen network options.

Option Setting Functionality/reasoning for selection

Set random
seed

Selected
and a spe-
cific seed
identified

As mentioned before, the training of the Kohonen
network model is initiated with random weights.
This setting allows the analyst to set a specific
value as a random seed for generating the initial
weights, ensuring reproducible results on different
runs
If this option is not selected, the model's result will
depend on the randomly selected seed, yielding,
to a certain extent, different results on different
runs, even if the other settings are held constant

Stop on Default Unless there is a strict time limitation, the model

training process should not be stopped before
concluding all the learning cycles and data passes.
This option ensures that data are passed through
the network an adequate number of times,
ensuring model refinement