How to create Utility Functions for dufferent types of parameters ...
You have not just prioritize the parameters, but also define which values of the Parameter are more preferable.
In other words, you have to assign a Utility Function for every Parameter.
Utility Function is a function, which maps the values of Parameter to the scale between 0 and 1.
Utility Function allows prioritize different values of the Parameter. The close Utility Function to 1 - the better
value of Parameter. Every type of the Parameter supposes specific view of the Utility Function.
1. Utility Function for Numeric parameters
You can assign a Utility function to the Numeric parameter using up to 5 points.
For every point you have to insert value of the parameter and then select corresponding value of the
Utility Function from drop-down list. To assign Utility function for numeric Parameter you have to describe
at least two points.
For example, if you are a doctor and you have to estimate condition of a patient you can use parameter "Temperature"
with the next Utility Function:
Point
Value (°C)
Utility Function
Description
Point 1
35.0
0.2
Mediocre
Point 2
36.6
1.0
Ideal
Point 3
37.5
0.3
Satisfactory
Point 4
42.0
0.0
Minimally Acceptable
For all the rest values of the Parameter the value of the Utility Function may be calculated. (We presume that for
all values less than minimal value in the table - less then 35.0 in our example - the value of the Utility Function
will be the same - 0.2. And similarly for all values more than maximum value the Utility Function will be the same.)
2. Utility Function for Logical (Yes/No) parameters
To assign a Utility Function to the logical Parameter you just have to answer a question - if this feature desirable
or not.
For example when you compare cars for the parameter "Air Conditioning" you can assign Utility Function
"This feature is desirable". It means that the value of the Utility Function equal to 1 for air-conditioned
cars, and equal to 0 for others.
3. Utility Function for Qualitative parameters
For qualitative parameters we have to assign value of the Utility Function to every value of the attached classifier.
You can do it in the same way as you assign weights to the parameters. In other words, you can give the values directly
or use pairwise comparisons.