Priors

Priors or Prior Probabilities are used when the sample data has a different distribution from that of the whole population. For example, suppose there are two target classes, responders and non-responders. Suppose that responders make up only 1% of population. For reasons of expense of data collection, convenience, or because of the rarity of the responders, the data might be collected such that it consists of 50% responders and 50% non-responders, that is, the responder population was over sampled. To allow the model to correct for this condition, you specify priors.

For each target value, adjust the Priors Distribution column so that it reflects the distribution in the population.

To restore the original display, click Restore.

When you are done, click OK to return to the previous dialog.

The wizard checks that you have specified valid priors values and reports any problems.

A mining activity may create priors for you.