Insurance risk score (DMN)
Scoring is where a Collect hit policy earns its place: several rules match, and their outputs are summed into a single score.
#What this example shows
A Collect hit policy with a sum aggregator.
Rules that deliberately overlap.
A score that is the total of every match.
#Columns
Column | Type and meaning |
|---|---|
Age (input) | number: applicant age |
Claims in 5 years (input) | number: count |
Vehicle group (input) | number: 1 to 50 |
Points (output) | number: contribution to the risk score |
#The decision table
Hit policy: C+ (Collect, sum). Cells are FEEL expressions, so a range, a list or a negation is a legal value; a dash means the input is irrelevant to that rule.
Age | Claims | Vehicle group | Points |
|---|---|---|---|
< 25 | - | - | 30 |
[25..70] | - | - | 0 |
> 70 | - | - | 15 |
- | 0 | - | 0 |
- | [1..2] | - | 20 |
- | > 2 | - | 45 |
- | - | <= 10 | 0 |
- | - | (10..30] | 15 |
- | - | > 30 | 35 |
#Making it your own
Collect without an aggregator returns a list. The plus sign is what makes it a sum.
Overlapping rules are correct here, which is why Unique would be wrong.
Keep the score-to-premium mapping in a separate decision.
#Related
The hit policy is the whole decision.
