Special Settings for Weighted Score Tables 

When you use the function type Weighted Score Tables, a model field weighting needs to be specified for each model field. Furthermore, a function in the model field needs to be specified so that a partial weighting can be assigned to each value of the field. The weighting of the model fields then determines the share of partial weightings that the score value has. This is calculated as follows: Score value (field1, field2, ...) = weighting1 x partial weighting1 (field1) + weighting2 x partial weighting2 (field2) + ...

How the partial weightings are determined depends on the type of model field:

·        With discrete model fields, you specify partial weightings for individual values of the field. You can determine an additional partial weighting that is assigned to the remaining values. This partial weighting is only applied if you selected the Treat as separate instance indicator during outlier handling. For more detailed information about outlier handling, see Settings for All Function Types below.

·        With continuous model fields, you enter partial weightings for individual threshold values. You also have to specify the function process for the partial weightings between the threshold values. You have the following options:

¡        One option is for the function process to be constant between the threshold values, in which case the partial weighting of the left or right threshold value (depending on the setting) is taken. In this case, the function process of the partial weighting is piecewise constant.

¡        Alternatively, the partial weighting is calculated continuously by linear interpolation.

¡        Furthermore, you also have the option of setting the partial weighting as equal to the value of the field.

In each case, you need to enter at least two threshold values because the value range applied in outlier handling (see below) lies outside of these values (that is, above the highest threshold value and below the lowest threshold value).

The score value to be defined is dependent on the discrete model field Customer Category and the continuous model field Income. The weighting of these two model fields should be 3 and 1 respectively. In the model field Customer Category, the data to be processed takes the values gold, silver, bronze, copper, and iron. The following partial weightings could then be specified:

Value	Partial Weighting
Gold	10
Silver	6
Bronze	4

 

The partial weighting 2 can be assigned to the remaining values. For the model field Income, the threshold values and corresponding partial weightings could be assigned as follows:

Threshold Value	Partial Weighting
0	0
10 000	10
25 000	20
50 000	30

 

The partial weightings function should be piecewise constant and take the partial weighting of the left threshold value in the interval between two threshold values. In this way, the score value (silver, 40 000) = 3 x 6 + 1 x 20 = 38 is obtained. If the partial weightings function for the income should be continuous instead of piecewise constant, then it produces the score value (silver, 40 000) = 3 x 6 + 1 x 26 = 44.

If the Treat as separate instance option was selected in outlier handling for the model field Customer Category, then the function produces the score value (iron, 10 000) = 3 x 2 + 1 x 10 = 16.

If the Constant extrapolation option was chosen in outlier handling for the model field Income, then the function produces the score value (silver, 60 000) = 3 x 6 + 1 x 30 = 48. If the Extrapolation option is chosen, this produces the score value (silver, 60 000) = 3 x 6 + 1 x 34 = 52.