Managing Uncertainty In Expert Systems

Managing Uncertainty in Expert Systems

3. Textbook for a course in expert systems,if an emphasis is placed on Chapters 1 to 3 and on a selection of material from Chapters 4 to 7. There is also the option of using an additional commercially available sheU for a programming project. In assigning a programming project, the instructor may use any part of a great variety of books covering many subjects, such as car repair. Instructions for mostofthe "weekend mechanic" books are close stylisticaUy to expert system rules. Contents Chapter 1 gives an introduction to the subject matter; it briefly presents basic concepts, history, and some perspectives ofexpert systems. Then itpresents the architecture of an expert system and explains the stages of building an expert system. The concept of uncertainty in expert systems and the necessity of deal ing with the phenomenon are then presented. The chapter ends with the descrip tion of taxonomy ofexpert systems. Chapter 2 focuses on knowledge representation. Four basic ways to repre sent knowledge in expert systems are presented: first-order logic, production sys tems, semantic nets, and frames. Chapter 3 contains material about knowledge acquisition. Among machine learning techniques, a methodofrule learning from examples is explained in de tail. Then problems ofrule-base verification are discussed. In particular, both consistency and completeness oftherule base are presented.

Managing Uncertainty in Expert Systems

A study of using probability to manage uncertainty in expert systems is presented. The study begins with a comprehensive summary of the literature on applying numeric techniques to manage uncertainty in expert systems. In addition to probability, fuzzy sets, certainty factors, and belief functions are addressed. basic principles and rules of information combination for each technique are discussed. The Lindley scoring rule argument for why probability is mathematically techniques is reviewed. The issues why using probability is considered to be a hindrance to managing uncertainty in expert systems are also reviewed. A simple expert system is developed using a state of the art expert system building tool called ALTERID. ALTERID is unique in that it unifies logical and probabilistic inference. This simple expert system is used to explore how probability theory can be used to manage the uncertainty in expert systems. The simple ALTERID based expert system is also used to evaluate the aforementioned issues for using probability to manage uncertainty in expert systems. Keywords: artificial intelligence Bayes theorem; decision analysis; theses.

Managing Uncertainty in Expert Systems