With Netica, decision nets are the same as regular Bayes nets, except that two additional types of nodes are present, decision nodes and utility nodes. In a regular Bayes net, all the nodes are called nature nodes because they have only to do with modeling the nature or reality of the world, the likelihood of its being in any of its possible states. The concept of utility and the concept of decision are seen as outside of merely depicting reality, being more in the realm of goals, desires, and agendas. Netica compiles both a Bayes net and a decision net into a junction tree for efficiency.
Tutorial: Use File → Open to read the decision net (also known as an influence diagram) "Umbrella" from the Examples folder. By opening a table dialog box for each nature node (beige oval) you can observe its conditional probability table, and by opening one for the utility node (green hexagon), you can see its utility function. If you look at the table dialog box for the decision node (blue rectangle), you will see that the node does not yet have any functional relation, which indicates that there is no decision function associated with the node. Our goal is to find a decision function which will maximize the expected value of the utility node.
This is accomplished by choosing Network → Compile. If you then look at the table dialog box for the decision node, it will display the discovered decision function (you will have to click the “Reset” button if you didn’t freshly open the dialog box). The numbers in the “Decide_Umbrella” give the expected utility of each choice. You can enter findings for Forecast to see how they change.
If you want to try another decision net, open “Car_Buyer_Neapolitan”, which is described in Neapolitan90, and is based on the classic “Car Buyer” influence diagram. It involves two sequential decisions about whether to do some tests and then whether to buy a certain used car. The optimal decisions turn out to be not to do any tests (D = none), but to buy the car (B = Buy when D = none).
Here
is a more detailed version of this exercise.