Combining DMN and the Knowledge Base Paradigm for Flexible Decision Enactment

Ingmar Dasseville, Laurent Janssens,

Gerda Janssens, Jan Vanthienen,Marc Denecker

DMN

Decision Requirements Diagram

Input

Output

U At Risk Model Convertible Price Potential Theft Rating
Boolean Boolean Currency (\$) {High, Moderate,Low}
1 True - - High
2 False True - High
3 False False >45K High
4 False False [20K...45K] Moderate
5 False False <20K Low

But wait... There's more!

```Information doesn't
do anything.

We do things with
information.```

Specifying Knowledge

1. What are you talking about?

2. What do you know about it?

``````vocabulary V {
type Price constructed from
{ lessthan20k
, between20And45k
, over45k}

type TheftRating constructed from
{ high
, moderate
, low}

Convertible
CarPrice : Price
HighTheftProbabibilityAuto
PotentialTheftRating : TheftRating
}``````

1. What are you talking about?

2. What do you know about it?

U At Risk Model Convertible Price Potential Theft Rating
Boolean Boolean Currency (\$) {High, Moderate,Low}
1 True - - High
2 False True - High
3 False False >45K High
4 False False [20K...45K] Moderate
5 False False <20K Low
``````PotentialTheftRating = high <- HighTheftProbabibilityAuto.
PotentialTheftRating = high <- Convertible.
PotentialTheftRating = high <- CarPrice = over45k.
PotentialTheftRating = moderate <- CarPrice = between20And45k
& PotentialTheftRating ~= high.
PotentialTheftRating = low <- CarPrice = lessthan20k
& PotentialTheftRating ~= high
& PotentialTheftRating ~= moderate.
``````

Specifying Knowledge

1. What are you talking about?

Be useful!

3. What is the required reasoning task?

2. What do you know about it?

Be useful!

3. What is the required reasoning task?

- Determining Consequences

- Expand to full solution

- Optimisation

Deriving Conclusion From Premises

- Verification

What If?

Handling Incomplete Data

Potential Theft Rating

demo

• Demonstrate the different reasoning tasks

The self-driving cars arrive!

Self-driving cars

demo

Modify the knowledge with an additional property

DMChallenge: Make a good burger

• <3000mg Sodium
• <150g Fat
• <3000 Calories
• Servings Ketchup = Servings Lettuce
• Servings Pickles = Servings Tomatoes
• At most 5 of each item

Make a good burger

demo

• Optimisation: Maximum Priced Burger
• Optimisation: Minimum Priced Burger

Profit

Constraint Programming (CP)

Satisfiability Modulo Theories (SMT)

Theorem Provers

...

Combining DMN and the Knowledge Base Paradigm for Flexible Decision Enactment

Ingmar Dasseville, Laurent Janssens,

Gerda Janssens, Jan Vanthienen,Marc Denecker

By krr

• 982