Inference engine
Complete inference engine can be divided into following functional elements:
- Control system - determines the order of testing the knowledge base rules
- Rules interpreter - defines a boolean (true, not true uncertainty factor) applications rules
- Explanation mechanism – justifies user the process of reasoning and generates report
A very important element of the expert system is also inference engine. Knowledge of the science must always be stored in the knowledge base, in formalized form, understandable to the inference engine. Using symbols you can easily determine how to handle the system to solve and analyze the correctness of the knowledge base. Since the inference engine is separated from the knowledge base it can be used in skeletal expert systems.
Reasoning comes with aplly of certain pattern, which allows tasks to be based on the veracity of premises, request that the truth of a different opinion being requested. It is an attempt to determine the truth of the hypothesis targeted by inference engine. It is based on the assumption that between sentences there is an objective inference ratio, or the ratio of probabilities. Inference requires the ability to make decisions based on their knowledge.
The inference is divided into reliable and unreliable.. An example of a reliable inference is deductive reasoning. A special variant thereof is syllogistic reasoning from two premises. Syllogism model contains major and minor premise, which show the application. The inference unreliable evidence does not warrant the truthfulness truth of the conclusion. The direction of this inference is considered to be inconsistent with the direction of logical consequence. Such inference is unreliable. Reductive inference. It is the inference method consisting in choosing the sentence recognized as true (the consequence) of such opinion (the reason), from which it follows logically first. Other examples of inference are unreliable: inductive reasoning by analogy.
In general inference can be written as a formula:
P1,P2,Pn - evidence
W - deduction
There are three basic types of reasoning: back (regressive), forward (progressive), and mixed. There are also methods of inference that use the uncertain knowledge . An example of this technique is called fuzzy logic. The most commonly used and most important methods of inference in expert systems are forward and backward chaining.
