Module 5
Expert Systems
Basic Concepts of Expert Systems
• 1. Introduction to Expert Systems
• Definition: Expert systems (ES) are computer-based information systems that utilize expert
knowledge to make high-level decisions in a specific problem domain.
• Notable Applications: MYCIN for medical diagnosis, tax accounting, credit analysis, environmental
monitoring, and fault diagnosis.
• 2. Experts and Expertise
• Experts: Individuals with specialized knowledge, judgment, experience, and skills in a particular field
who can provide solutions to complex problems.
• Expertise: The extensive, task-specific knowledge that experts have, typically acquired through
training, reading, and experience.
• Characteristics of Expertise:
• Often associated with high intelligence, but not necessarily the smartest person.
• Based on vast amounts of knowledge and the ability to learn from experience.
• Knowledge must be well-organized and quickly retrievable.
3. Features of Expert Systems
• Expertise Representation: ES must possess expertise akin to human
experts to make decisions.
• Symbolic Reasoning: The primary reasoning mechanism involves symbolic
representation of knowledge and symbolic reasoning mechanisms like
backward and forward chaining.
• Deep Knowledge: ES contains complex, deep knowledge that non-experts
typically do not have.
• Self-Knowledge: ES can examine their reasoning, update knowledge
constantly, and learn from successes and failures.
4. Development and Generations of Expert Systems
• First-Generation ES: Use if-then rules to represent and store knowledge.
• Second-Generation ES: More flexible, integrating multiple knowledge
representation and reasoning methods such as fuzzy logic, neural
networks, and genetic algorithms.
Applications of Expert Systems
[Link] Diagnosis: Systems like MYCIN used for diagnosing diseases and
suggesting treatments.
[Link] Analysis: Evaluating credit applications and determining
creditworthiness.
[Link] Accounting: Assisting in tax preparation and compliance.
[Link] Monitoring: Analyzing environmental data for monitoring and
prediction.
[Link] Diagnosis: Identifying and diagnosing faults in machinery or systems.
[Link] Assistance: Providing legal advice and case analysis.
[Link] Support: Automated customer service solutions.
[Link] Planning: Assisting in personal and corporate financial planning.
[Link] Management: Helping with crop management and pest control.
[Link] Exploration: Analyzing geological data for oil exploration.
•Knowledge Acquisition Subsystem: Involves the transfer and
transformation of expertise from human experts to the ES.
•Knowledge Base: Stores refined and organized knowledge for
future reference.
•Inference Engine: Uses knowledge rules and data to derive
conclusions and solutions.
•User Interface: Facilitates interaction between the user and the
system.
•Blackboard: Serves as a workspace for storing intermediate
results and facilitating collaborative problem-solving.
•Explanation and Knowledge-Refining Subsystems: Ensure
transparency and continuous improvement of the ES.
Knowledge Engineering
• Definition and Purpose:
• Knowledge Engineering encompasses the acquisition of knowledge
from human experts or other information sources and its conversion
into a repository known as a knowledge base. This process supports
the development of systems capable of solving complex problems
requiring expert knowledge.
• Activities in Knowledge Engineering:
[Link] Acquisition:
1. Involves obtaining knowledge from experts, books, documents, sensors, or
computer files.
2. Knowledge can be specific to a problem domain or general problem-solving
procedures. This includes metaknowledge (knowledge about knowledge).
2. Knowledge Representation:
•Organizes acquired knowledge in a format that is usable for problem-solving.
•Represents a knowledge map encoding the knowledge base's contents.
3. Knowledge Validation:
•Involves verifying the acquired knowledge's accuracy and quality using test
cases.
•Ensures the knowledge is appropriate for use in expert systems.
4. Inferencing:
•Designing software that uses stored knowledge to make inferences and solve
problems.
•The system can provide advice to non-experts based on stored knowledge.
5. Explanation and Justification:
•Programming the system to explain why certain pieces of information are
needed or how conclusions were derived.
•Enhances the system's transparency and trustworthiness.
Process of Knowledge Engineering:
• Figure 12.5 outlines the steps: starting from problem identification,
through knowledge acquisition, representation, validation, inferencing,
and explanation, forming a feedback loop for continuous improvement.
• Challenges in Knowledge Acquisition:
[Link] Difficulty:
1. Experts may struggle to articulate their knowledge or be unwilling to share it.
[Link] Constraints:
1. Experts may lack the time to participate in knowledge acquisition processes.
[Link]:
1. Testing and refining acquired knowledge can be complex and resource-
intensive.
[Link] Representation:
1. Difficult to represent specific knowledge accurately, especially when mixed with
irrelevant data.
Types of Knowledge in Knowledge Engineering:
• Theories and Models: About the problem domain.
• Rules and Procedures: Specific to the problem domain.
• Heuristics: General problem-solving strategies.
• Global Strategies: For various problem types.
• Metaknowledge: Knowledge about the knowledge process.
• Facts: About the problem area.
• Knowledge Verification and Validation:
• Evaluation assesses the overall value and accuracy of the expert
system.
• Verification ensures the system performs correctly according to
specifications.
Knowledge Representation Methods:
• Production Rules: Represent knowledge as conditions and actions.
• Semantic Networks, Frames, Objects: Various other forms of
knowledge representation suitable for different scenarios.
• Inferencing Mechanisms:
• Forward Chaining: Data-driven approach that works from available
information to conclusions.
• Backward Chaining: Goal-driven approach that works from goals to
data needed to achieve them.
• Example of Inferencing:
• An investment decision-making scenario demonstrates how rules are
applied to make decisions based on given data, showcasing both
forward and backward chaining methods.
Problem Areas Suitable for Expert Systems
• Expert systems (ES) can address a wide range of problem areas by inferring situation descriptions
from observations. The following are the generic categories of ES along with the problems they
address:
1. Interpretation Systems
1. Description: These systems infer situation descriptions from observations.
2. Applications: Surveillance, speech understanding, image analysis, signal interpretation, and intelligence
analysis.
2. Prediction Systems
1. Description: These systems infer likely consequences of given situations.
2. Applications: Weather forecasting, economic forecasting, traffic predictions, crop estimates, demographic
predictions, and military and financial forecasting.
3. Diagnostic Systems
1. Description: These systems diagnose system malfunctions from observations.
2. Applications: Medical, electronic, mechanical, and software diagnoses, typically based on observed
behavioral irregularities.
4. Design Systems
1. Description: These systems develop configurations of objects that satisfy the constraints of a design problem.
2. Applications: Circuit layout, building design, and plant layout.
5. Planning Systems
1. Description: These systems specialize in planning problems, developing plans to achieve goals.
2. Applications: Automatic programming, project management, routing, communications, product
development, military applications, and financial planning.
6. Monitoring Systems
1. Description: These systems compare observations of system behavior with standards to flag deviations.
2. Applications: Air traffic control, fiscal management tasks, and other topics requiring real-time monitoring.
7. Debugging Systems
1. Description: These systems diagnose problems and recommend solutions.
2. Applications: Planning, design, and prediction capabilities to correct diagnosed problems.
8. Repair Systems
1. Description: These systems develop and execute plans to administer a remedy for diagnosed problems.
2. Applications: Incorporate debugging, planning, and execution capabilities for various domains.
9. Instruction Systems
1. Description: These systems diagnose and correct student performance, constructing a description of the
student's knowledge.
2. Applications: Educational settings to identify and remedy deficiencies in student knowledge.
10. Control Systems
1. Description: These systems adaptively govern the overall behavior of a system by interpreting current
situations, predicting future states, diagnosing problems, formulating remedies, and monitoring
execution.
2. Applications: Continuous system monitoring and control in various industrial and operational settings.
Development of Expert Systems
• Defining the Nature and Scope of the Problem
• Initial Step: Identify the nature of the problem and define its scope.
• Suitability: Not all problems are appropriate for expert systems (ES). For
example, problems that can be solved by mathematical algorithms might
not be suitable for ES.
• Characteristics: Qualitative problems are more suited for rule-based ES
where expertise is explicitly provided.
• Feasibility: The scope should be specific and reasonably narrow to ensure
effective problem-solving using available technology.
• Identifying Proper Experts
• Expert Identification: After defining the problem, the next step is to find
experts willing to assist in developing the knowledge base.
• Expert Criteria: Experts should have a thorough understanding of problem-
solving knowledge, decision support technology, and good communication
skills.
• Acquiring Knowledge
• Knowledge Engineering: The process of eliciting knowledge from experts is
called knowledge engineering.
• Complexity: Knowledge acquisition is time-consuming and can be hindered
by experts' unwillingness to share proprietary knowledge or time
constraints.
• Knowledge Representation: Knowledge is documented in a form suitable
for use in the ES, often as if-then rules.
• Selecting the Building Tools
• General-Purpose Development Environment: Tools like C++, Prolog, and
LISP, used for developing ES.
• Expert System Shells: These are specialized tools with built-in inference
capabilities and user interfaces but empty knowledge bases.
• Tailored Turn-Key Solutions: Custom solutions designed for specific
domains, providing a quick adaptation of ES tools to particular applications.
• Coding the System
• Implementation: The development team codes the knowledge based
on the tool's requirements, ensuring the coding process is efficient
and error-free.
• Evaluating the System
• Verification: Ensures that the ES contains knowledge exactly as
acquired from the expert, solving the problem accurately.
• Validation: Checks whether the knowledge acquired from the expert
can effectively solve the problem.
• Benefits, Limitations, and Critical Success Factors
• Benefits:
• Increased Output and Productivity: Faster performance compared to humans.
• Decreased Decision-Making Time: Quick recommendations and decisions.
• Increased Process and Product Quality: Reduced errors and improved quality.
• Reduced Downtime: Efficient diagnosis and repair of machine malfunctions.
• Operation in Hazardous Environments: Can work in unsafe conditions for humans.
• Accessibility to Knowledge: Provides valuable insights and advice.
• Ability to Work with Incomplete Information: Handles uncertain data effectively.
• Provision of Training: Acts as a teaching tool for novices.
• Enhanced Problem Solving: Integrates expert judgment into the decision-making
process.
• Improved Decision Quality: Provides consistent and reliable decisions.
Limitations:
•Knowledge Acquisition: Difficult to extract expertise from humans.
•Narrow Domain: Limited to specific knowledge areas.
•Dependence on Experts: Requires expert validation and
verification.
•High Development Costs: Expensive and time-consuming to
develop.
•Potential for Incorrect Recommendations: Human experts may
need to resolve issues.
Critical Success Factors:
•Knowledge Sufficiency: Adequate knowledge from experts.
•Cooperation: Strong support from management and experts.
•System Design and Usability: User-friendly interface and effective
implementation.