Expert systems Revision Notes

Expert systems are specialised computer programs that mimic the decision-making abilities of human experts in specific domains. Unlike general-purpose software, they focus on solving complex problems by applying expert knowledge and logical reasoning.

These systems are designed to assist users by providing advice, diagnoses, or solutions based on input data. They are particularly useful where expert human knowledge is scarce, expensive, or inconsistent.

For example, an expert system in medical diagnosis can analyse symptoms entered by a doctor and suggest possible illnesses, helping to improve accuracy and speed of diagnosis.

• Simulate human expert reasoning.
• Provide advice or solutions in specialised areas.
• Useful when expert knowledge is limited or costly.

An expert system consists of five main components that interact to process information and generate conclusions: the user interface, inference engine, knowledge base, rules base, and explanation system.

The user interface allows users to input data and receive advice or solutions. It must be intuitive to facilitate effective communication between the user and the system.

The knowledge base stores domain-specific facts and information, such as medical symptoms or mineral properties. This data forms the foundation for reasoning processes within the system, but on its own, it does not provide conclusions without rules to interpret it.

• User interface: facilitates user interaction.
• Knowledge base: contains domain facts and data.
• Rules base: holds conditional logic guiding decisions.
• Inference engine: applies rules to knowledge to infer conclusions.
• Explanation system: clarifies reasoning behind results.

The inference engine is the core processing unit that applies logical rules from the rules base to the facts stored in the knowledge base. It uses these rules, often structured as IF-THEN statements, to deduce new information or reach conclusions.

For example, in diagnosing car engine problems, a rule might be: IF the engine fails to start AND the battery is charged, THEN check the fuel system. The inference engine evaluates such rules against the input data to suggest possible causes or solutions.

This reasoning process can be forward chaining (data-driven) or backward chaining (goal-driven), depending on how the system searches for conclusions.

• Inference engine applies IF-THEN rules to knowledge base facts.
• Rules base contains conditional statements guiding decisions.
• Enables logical deduction of new information.
• Supports forward and backward chaining reasoning methods.

The explanation system provides transparency by explaining how the expert system arrived at a particular conclusion or recommendation. This is crucial for user trust and understanding, especially in critical fields like medicine or finance.

For instance, if an expert system suggests a diagnosis, the explanation system can show which symptoms and rules led to that conclusion, allowing users to verify and learn from the process.

This feature distinguishes expert systems from simple databases or decision trees by offering insight into the reasoning process.

• Explains reasoning behind conclusions.
• Increases user trust and understanding.
• Shows which rules and facts influenced decisions.

Expert systems are applied in diverse fields where expert knowledge is valuable and complex decision-making is required.

Examples include: mineral exploration, where systems analyse geological data to identify promising sites; diagnosing car engine faults by evaluating symptoms and sensor data; medical assessments to suggest possible illnesses based on patient symptoms; chess strategy, where systems evaluate board positions and suggest moves; financial planning by analysing market data and personal finances; scheduling delivery routes to optimise logistics; and identifying plants and animals using characteristics and classification rules.

These applications demonstrate how expert systems can improve efficiency, accuracy, and accessibility of expert knowledge.

• Mineral exploration: analysing geological data.
• Car engine diagnosis: identifying faults from symptoms.
• Medical assessments: suggesting possible illnesses.
• Chess strategy: recommending moves based on board state.
• Financial planning: advising on investments and budgets.
• Delivery route scheduling: optimising logistics.
• Plant and animal identification: classifying species from traits.