Rete_Rule_Engine.pdf

Transcript

1. Rete Rule Engine Ganesh Gunasegaran https://itsgg.com me@itsgg.com
2. What? • Pattern matching algorithm for matching facts against the patterns in rules.
3. Overview
4. Concepts: Rules • Represent some knowledge. • In the form of if-then-else. • If: one or more condition. • then: one or more action.
5. Concepts: Facts • Facts are data. • Number of facts increases with the number of unique variables.
6. Concepts: Ruleset • Knowledge base consisting of one or more rules.
7. How? • A complete rule-set is given to the rule engine. • The rule engine matches each rule in the ruleset with given facts. • Decide whether to fi re the rule or not. • Repeat!
8. How? - Cont. • New facts may be added to the list. • Old facts may be removed from the list. • Actions of the rules can change a few facts. • Remembers what has already matched and computes only the changes.
9. Inference Cycle
10. Phase 1 - Match • The conditions of the rules are matched agains the facts to determine which rules are to be executed. • The rules whose conditions are met are stored in a list called Agenda for fi ring.
11. Phase 2 - Select • The selection criteria depends on. • Priority • Recency of usage • Speci fi city of the rule
12. Phase 3 - Execute • The selected rule is executed by carrying out the actions in the right hand side of the rule . • The action can be • An assertion • A user de fi ned function • A built-in function • A decision table
13. Internal working • A directed acyclic graph consisting of nodes representing patterns in the conditions of the rules is created. • The nodes behave like fi lters, testing the incoming tokens, and sending only those that pass the test.
14. Internal working - Nodes • Alpha Network • Each alpha node has one input that de fi ne intra-elements. • Beta Network • Each beta node takes two inputs to de fi ne intra-elements.
15. Internal working - Example • if age > 60 or age < 5 or income < 10, 000 then concession = 50%
17. Advantages • Speed as it takes the advantage of structural similarity in rules. • Pools the common components so that they need not be computed again.
18. Disadvantages • Poorly written rule can be memory intensive. • The space complexity is O(RFP) • R -> Number of rules • F -> Number of asserted facts • P -> Number of patterns per rule
19. That’s all folks!