Integrate a predictive model in a decision model to enrich your decision
making.
About this task
Predictive models are meant to be reused in decision models to make decisions based on
predictions. When you design the decision model, you must think about the part of the decision logic
where you want to use the predictive model. Then, you configure the corresponding decision node to
use the predictive model.
Procedure
-
Open the decision model where you want to integrate the predictive model.
- In the diagram toolbar, click Add prediction to create a
prediction node.
- Configure the node by selecting the predictive model that you want to use.
- In the diagram, create a dependency between this node and the decision node that will use
the prediction to make a decision: hover over the prediction node, click the Connect to
another node button
, and select the decision node.
- When you write the decision logic of the decision node, use the output of the predictive
model to make your decision.
For example, if you want to define interest rates based on the risk score computed by a
predictive model, you can create a decision table in which the output of the predictive model is
used to define a condition column:
| |
Loan duration |
Predicted risk score |
Annual interest rates |
| Min |
Max |
Min |
Max |
| 1 |
< 5 |
0 |
0.7 |
0.05 |
| 2 |
< 5 |
0.7 |
0.8 |
0.052 |
| 3 |
5 |
8 |
0 |
0.7 |
0.056 |
| 4 |
5 |
8 |
0.7 |
0.8 |
0.057 |
| 5 |
9 |
12 |
0 |
0.7 |
0.06 |
| 6 |
9 |
12 |
0.7 |
0.8 |
0.061 |
Where the
Predicted risk score condition column is defined
as
follows:
the risk score computed from
Borrower being Borrower ,
Loan being Loan ,
is at least <min> and less than <max>
Tip: The invocation example available in the Details tab of the
function node can help you write the decision logic.
What to do next
Run the decision model to verify that it behaves as expected with
the prediction.