A methodology for uncertainty in knowledge-based systems (Q1188646)

This research monograph presents a methodology for representing and handling uncertainty in expert systems; especially medical diagnostic systems. The methodology is founded on classical probability theory, but takes into account that in many expert systems the available information is too weak to produce reliable point estimates for probability values. Therefore the handling of interval-valued probability is one of the main goals of the monograph. The monograph includes discussions of other schemes for uncertainty management, such as MYCIN and Dempster-Schafer. An argument is made that the Dempster-Schafer combination rules can produce misleading results. Similarly negative conclusions are drawn for the MYCIN's approach. The monograph's treatment is rigorous and covered with many examples. The examples are rather simple. The monograph does not discuss all relevant recent literature on the subject of handling uncertainty in expert systems. There is a definitely growing interest in the topic of interval probability theory, and this monograph represents a good contribution.