DSS & Warehousing Systems Chapter 11 Efrem Mallach Prepared by Luvai Motiwalla Irwin/McGraw-Hill Copyright © 2000 by The McGraw-Hill Companies, Inc. All rights reserved.
Expert Systems Introduction Expert systems and DSS Confidence factors Fuzzy logic Expert system development tools Choosing a good expert system application Finding the expert(s) Pros and cons of expert systems
Introduction Artificial Intelligence:The field of artificial intelligence encompasses several areas,of these expert systems are the most useful in DSS overall. Pages 426 to 427Artificial Intelligence:The field of artificial intelligence encompasses several areas, including robotics, machine vision, understanding languages , speech recognition, neural networks, and expert systems. Of these expert systems are the most useful in DSS overall.
Expert Systems and DSS Expert systems are not an alternative to DSS. Rather, expert system,s are a technology that is often useful in constructing a DSS. Pages 427 to 430 Expert systems are also called knowledge – based systems or rule - based systems. Information systems apply the reasoning process of a human expert to new situations. Such systems are usually are especially useful in decision making.
Confidence Factors What are confidence factors? Where confidence factors come from? Pages 430 to 432 Expert systems handle situations by using confidence factors or truth values. A confidence factor is a value associated with a variable that indicates the degree to which the system treats that variable as true. An expert system can obtain a confidence factor to associate with a piece of knowledge in one or more of three ways: A user can enter the confidence factor to be associated with an input value. A confidence factor can be obtained from a database. A rule can incorporate a confidence factor into a logic system. Confidence factors can be propagated forward from one rule to another.
Fuzzy Logic Fuzzy logic is an approach to dealing with imprecise knowledge in a precise way. Pages 432 to 434 Fuzzy logic provides a rigorous mathematical method of dealing with subjectively defined terms. Fuzzy logic – based expert systems can use modifiers to qualify relationships further. The advantages of fuzzy logic for expert systems is that it matches the way human experts think about many important problem areas.
Expert System Development Tools- cont’d Expert systems are usually developed using preprogrammed shells. A shell provides an inference engine and a user interface. Pages 434 to 440 The system developer need only write the rules in the language that the inference engine understands. If it is necessary to write an expert system from scratch, without a shell, the specialized languages of LISP and Prolog have been developed specifically for artificial intelligence applications.
Expert System Development Tools Languages: There are several reasons one might develop an expert system without using a commercial shell. Pages 440 to 442 These reasons are : No commercial shell meets requirements. To learn about expert systems by building one from the ground up. The expert system is a small part of a larger application. The programmer already knows another language and doesn’t want to learn a shell. Because the resulting system must be across multiple platforms, which do not all support the same shell or compatible shells.
Choosing a Good Expert System Application-cont’d It is convenient to divide the selection characteristics into (A) the characteristics of the problem to be solved Pages 442 to 447 Problem – related ( task --related ) criteria The task should typically take human experts from several minutes to a few hours to solve. The task must involve only processing information and cognitive skills. The task must be carried often. The task must be reasonably stable. The task must involve only explicitly visible knowledge. The required knowledge must be within an acceptably narrow area. The necessary data must be available to the system. The task must have financial importance to its sponsor. Test cases must be available. We should be able to tolerate errors in system output.
Choosing a Good Expert System Application (B) the characteristics of the human experts who solve the problem. Pages 445 to 447 Expert – related criteria: We must be able to tell who the experts are. Experts must perform the task substantially better than non experts. One or more experts must be available to work on the system. the human expertise must be scarce. Experts must agree on the solutions.
Finding the Expert(s) At least one competent, willing, and cooperative expert must be available to develop an expert system. Pages 447 to 449 Good experts have the following characteristics: They must be experts- people who are up to date. They must be willing to be cloned by the computer. they must be able to spend substantial time on the project. They should be personally cooperative and easy to work with. They must be interested in computers or at least intrigued by the technology.
Expert Systems and DSS Expert systems reproduce the reasoning process a human decision maker would go through in reaching a decision, diagnosing a problem or suggesting a course of action. Pages 449 to 451 Expert systems can also incorporate a database and models,as used in other types of DSS. The ability to explain its reasoning is inherent in the knowledge and rule oriented structured of an Expert system. Other types of DSS either do not have this ability or must have it explicitly added to them via additional program code.
Pros and Cons of Expert Systems Expert systems can provide substantial benefits in some DSS. At the same time, no tool is perfect for all applications.It always has its advantages and drawbacks. Pages 451 to 452 Advantages of expert systems: They solve problems more quickly than humans. Their output is consistent. They can be replicated as needed with minimal lead time and at moderate cost. They don’t cost money when they are not being used. They can free human experts to other tasks. They don’t mind working in locations that people find convenient or hazardous, or taking repeated tasks that people find boring. They can be expanded by adding more rules. They can train novice problem solvers. Drawbacks of expert systems: Their domain of expertise is usually narrow. They cannot apply common sense, only their rules. They are brittle at their limits.They may be costly to develop. One or more expert must be on hand to contribute to the project for an extended period of time.