Introduction to Ranking Models - Text Mining Techniques for Healthcare Provider Quality Determination - page 23

Once the “Died” variable has been changed to binary, we can investigate the decisions (Figure 9),
and the importance of each decision.
We want to build the decisions, including prior probabilities and weights. The prior probabilities are
those that exist in the population (Figure 10). In this example, mortality is a rare occurrence (approxi-
mately 2% of all patients in the dataset). A discussion of how to accommodate a rare occurrence with
these prior probabilities is given in Chapter 4.
One of the things we need to consider is whether we want to weight a false negative differently from
a false positive in terms of cost. While this will be discussed in more detail in Chapter 4, we want to
Figure 10. Prior probabilities for target variable (created with SAS® software. Copyright 2009, SAS
Institute Inc., Cary, NC, USA. All Rights Reserved. Reproduced with permission of SAS Institute Inc.,
Cary, NC)
Figure 11. Decision weights (created with SAS® software. Copyright 2009, SAS Institute Inc., Cary, NC,
USA. All Rights Reserved. Reproduced with permission of SAS Institute Inc., Cary, NC)