RETE Algorithm
Used in forward chaining rule-based systems.
- Performs very well in situations in which very few changes occur (i.e. where there are few new rules or changes to rules).
- Has significant storage overhead.
- The RETE algorithm and its derivatives are the only algorithms for implementing production (inference) systems whose performance is independent of the number of rules.
A hybrid system can combine both the backward-chaining and forward-chaining approaches to inference. Many rules-based systems used for business rules modelling are hybrid systems.
The purpose of the inference approach is to maximise the utility of the knowledge gained by the expert system. The utility of the knowledge is enhanced by using inference to allow rules to be applied to situations that were not necessarily foreseen at the time the system was developed. Actors that use the system can then use this knowledge to help solve problems in a particular domain, without the system having to be designed around the particular problems that are to be solved.