Custom Costs
Use this text box to specify a custom cost function describing costs of classification for binary or nominal dependent variables.
When you specify a custom cost, the predicted class is the one with the smallest expected cost according to the cost assignments that you make and the computed predicted probabilities.
By default, any misclassification is assigned a cost of 1 and any correct classification is assigned a value of 0. You need to specify only the costs that are different from these defaults.
To Specify a Custom Cost:
Costs are specified using a text string with the operators "as" and "=" to define the classification. Different classifications are separated by commas.
8 | Type the cost statement in the text box. |
For example, for a binary dependent variable with values A and B, specify B as A = 0.5, A as A = -0.5 (shown below).
If the dependent variable has numeric values 0 and 1, an example specification would be 0 as 1 = 0.75.
8 | If you change your mind and would like to reset the field, click . |