Expert Systems Infsy 540 Dr Ocker Expert Systems

Expert Systems Infsy 540 Dr. Ocker

Expert Systems n computer systems which try to mimic human expertise n produce a decision that does not require judgment assistants to decision makers rather than substitutes for them n

Expert Systems & AI n artificial intelligence (AI) - group of technologies that attempt to emulate certain aspects of human behavior, such as reasoning and communicating n Expert systems are the most important product of AI research to date

Systems and Types of decisions Traditional computing systems n deal with routine/structured problems e. g. payroll system is structured – – can write down the formulas calculations are very repetitive requires little judgment no creativity

Systems and Types of decisions DSS n deal with semi-structured problems – processing less well-defined – no judgment inside system – decision maker applies assumptions to problem, system calculates results – output must be interpreted by decision maker

Example Problem Use DSS to help predict effects of early retirement plan n vary assumptions about plan to forecast financial impact n system produces an answer for each set of assumptions n user judges the validity of the assumptions and the value of the answer

Systems and Types of decisions Expert System n deals with semi-structured problems – judgment incorporated into system – system produces a solution – system can “explain” how it reached its conclusions

Example Problem Use Expert System to develop an early retirement plan n system contains decision criteria (“rules”) established by decision makers n uses rules to frame a retirement program n can trace rules used in developing the retirement program

What is an expert system? A knowledge-based system: n provides specific knowledge about a narrow problem domain n knowledge stored in the knowledge base n system uses knowledge and an inferencing (reasoning) procedure to solve problems that would otherwise require human competence or expertise.

To use an expert system (1) gather input problem variables and criteria (2) consult computerized base of knowledge (3) system reasons out an answer ES often assistants to decision makers and not substitutes for them i. e. use ES to help DM with part of a larger problem

Example - Internist/Caduceus n n one of most knowledge-intensive expert systems covered 85% of internal medicine included information on 500 diseases and more than 100, 000 symptomatic associations user inputs given patient information system uses its knowledge base to identify a disease and recommend treatment

Components of expert systems 1) 2) 3) 4) knowledge base inference engine knowledge acquisition module explanatory interface

1. knowledge base structure for saving facts and rules relevant to a specific application (problem domain) 2 types of info: n (1) book knowledge about a domain n (2) heuristic knowledge - rules of thumb used by human experts who work in the domain n

2. inference engine n n that portion of the sw that contains the reasoning methods expert system asks questions of the user to get info. it needs. then inference engine, using knowledge base, searches for the sought-after knowledge returns a decision/ recommendation to user

3. knowledge acquisition module n used by expert to enter rules or facts into the system

4. explanatory interface system shows the trail of reasoning it used to reach a decision n explains the facts it used n what rules it applied n and in what order

2 environments of ES n n Development Environment Consultation Environment

Knowledge Representation knowledge represented in expert systems in variety of ways, including: n rules n case-based reasoning

Rules n n most common way to represent knowledge in expert system rules called heuristics - obtained from experts number of rules determines complexity of system rules most appropriate when knowledge can be generalized into specific statements.

example of heuristic rule if good customer and credit requested < $5, 000 and loan term < 1 year then grant credit

Case-based reasoning n n system draws inferences by comparing a current problem (case) with hundreds/thousands of similar past cases. best used when situation involves too many nuances and variations to be generalized into rules

example of case-based reasoning n Sharon , 35 yrs. old, entered hospital with potentially fatal respiratory disease. Her vital stats. and medical history entered into expert system. System drew on records of over 17, 000 previous intensive-care patients to predict whether Sharon would live or die.

example of case-based reasoning n n n First prediction - 15% chance of dying. Stats. entered daily - system compared her progress to base of previous cases. 2 weeks later - prediction soared to 90% chance of dying - alerted physicians and nurses to take corrective action.

How expert systems work knowledge representation n method used to organize knowledge n production rules - most common method – consist of an IF part and a THEN part n IF condition THEN action

How expert systems work Inference Engine n controls the order in which the production rules are applied to solve the problem and n resolves conflicts if more than one rule applies – this is the "reasoning" process

How solution process works n n user presents a set of facts describing a situation to the expert system. inference engine compares facts of the case to the knowledge base system then gives a recommendation asks for more information if needed

Inferencing strategies for rulebased system n Forward chaining – data driven – inferencing moves from facts of case to a goal (conclusion) n Backward chaining – inferencing moves from a possible goal state to premises that would satisfy it

Expert system shells n n Most common way to develop ES shell is an expert system without the knowledge base – includes inference engine, user interface, explanation and knowledge acquisition pieces – generic shells - used to develop ES in any domain – domain-specific shells - incomplete specific ES; require much less effort - already includes many rules

Development of ES n Prototype-oriented iterative development

Benefits of Expert Systems n n Quick consistent low error rate capture scarce expertise

Limitations of Expert Systems n n Must have agreement among experts must have a willing expert most only support operational-level tasks use can weaken human expertise

Appropriate Problem Space for Expert System 1. technical disciplines with large bodies of complex information 2. situations that require decisions 3. an expert can articulate the decision rules s/he uses