Expert System Knowledge Base Builder

CSC-266 · Semester IV · Artificial Intelligence

intermediate30 minAvailable

Expert System Knowledge Base

Toggle patient symptoms, then forward-chain the rules.

Facts (toggle)

Patient has fever Patient has cough Patient has rash Patient has sore throat

Rule base

IF fever AND cough THEN likely flu

IF fever AND rash THEN likely measles

IF flu THEN recommend rest and fluids

IF measles THEN recommend isolation

Derived so far

none — toggle some facts

In forward chaining, satisfied rules fire and their conclusions become new facts, cascading until nothing new can be derived (e.g. facts fever+cough → flu → rest).

Lab brief

Objective

Design a small rule-based knowledge base for a diagnostic expert system and test it against a set of patient cases.

Theory & concepts

An expert system stores knowledge as facts and IF-THEN production rules. Given the facts entered for a case, the inference engine fires rules whose premises are satisfied and adds their conclusions to the knowledge base. The system's answer is the chain of rules that led to a diagnosis.

Rendering diagram…

Facts and rules form the knowledge base of an expert system

RULE flu_likely:
  IF fever AND sore_throat AND NOT runny_nose
  THEN diagnosis = flu

RULE cold_likely:
  IF runny_nose AND cough
  THEN diagnosis = cold

Designing good rules

Rules work best when each premise is a fact the user can supply or another rule can conclude. Avoid conclusions that are themselves premises of the same rule, and check for rules that can never fire because no fact or rule produces their premise.

Prerequisites

Lesson: Expert systems
Lesson: Knowledge representation schemes

Instructions

1Open the rule editor and review the starter rules for the diagnostic domain.
2Add at least two of your own rules covering a new conclusion.
3Load a sample case and run the inference to see which rules fire.
4Examine the explanation trace showing the fact-to-conclusion chain.
5Test a case that matches none of your rules and refine the knowledge base.

Tasks

Extend the rule base
Add rules that conclude a new diagnosis and confirm they fire on a matching case.
Trace a conclusion
Run a case and read the explanation chain for the diagnosis reached.
Find the coverage gap
Construct a case with no matching rule and describe what the system reports.
Audit a rule
Identify a rule that fires incorrectly and adjust its premise.

Self-check

Each added rule fires exactly on cases that satisfy its premises.
The explanation trace correctly lists the rules used.
Unmatched cases are reported without a false conclusion.
The student can debug a rule by adjusting premises.

Environment

rule editorfact entryinference runexplanation tracecase library