The mixture of incomplete reasoing with inexact reasoning and its application to a real expert system

"Incomplete reasoning, is such a kind of reasoning that we could gain a conclusion with some premises omitted. The conclusions thus acquired are, in fact, conjectures. They need to be verified or refuted. We call the verification or refutation process "regression". In contrast to the complete reasoning of formal logic, incomplete reasoning tallys with our habit of reasoning more closely. In fact, studies show that any reasoning process is a kind of incomplete reasoning. In this paper, we construct an incomplete reasoning model (AIR model) for expert systems, utilizing the chniques of inexact reasoning. We give a practical AIR model basing on MYCIN's CF theory, which is implemented in a real expert system, i.e. Diesel Failure Diagnosis (DFD) expert system. Studies show that the incomplete reasoning makes the reasoner focus his (its) attention, and so raises the efficiency of reasoning. The mixture of incomplete reasoning with inexact reasoning in expert systems could deal with not only the inexactness of reasoning introduced by expert knowledges or initial evidences, but also the incompletity of reasoning caused by system's active conjecturing.